n8nNeed a weekly list of fresh PubMed papers
The lab gets a curated 10-paper reading list in Slack before the weekly meeting — no manual search needed.
Track 6 · Advanced
6.1
Automation tools
Wire up apps and data pipelines — pick the right one
1Overview
The no-code tools that connect your apps and automate work: n8n (self-hostable workflows), Make and Zapier (visual app automation), KNIME (visual data science), and Dify (LLM apps). Compare them as cards, switch to list view for the full cheat sheet.
1.1After this chapter you can
→Pick the right automation tool for a given task
→Know when to self-host (n8n/KNIME) vs use a hosted service (Make/Zapier)
→Tell data-processing tools apart from app-integration tools
1.2Which tool is fastest for popular apps?
Zapier offers a hosted drag‑and‑drop builder that lets you connect the most common SaaS apps in minutes, so you can get an integration up and running faster than any self‑hosted option.
1.3How can I host my own workflows?
Both n8n and KNIME can be installed on your own servers, giving you full control over data, code extensions, and execution while letting you design visual multi‑step flows without leaving your infrastructure.
1.4What tool handles visual data tables?
KNIME provides a visual environment for cleaning, joining, and analyzing tabular data, making it ideal when you need to manipulate spreadsheets or databases as part of an automated pipeline.
2Matrix 8 rows · 5 tools
knime
Self-host / own your data
yes
no
no
yes
yes
Free to run
yes
partial
partial
yes
yes
Drop into code
yes
partial
no
yes
partial
App integrations
many
lots
most
few
few
Built-in AI / LLM
yes
partial
partial
partial
yes
Data / tabular crunching
partial
partial
no
yes
no
Scheduling & triggers
yes
yes
yes
partial
partial
Beginner-friendly
partial
yes
easiest
partial
partial
3Sub-chapters
- 6.1.1 n8n Open-source workflow automation — the lab bench of this course 210
- 6.1.2 Make Hosted visual canvas — branches, loops and error handlers you can see 23
- 6.1.3 Zapier The largest app catalogue — the fastest first automation 17
- 6.1.4 Dify Open-source LLM app builder — upload papers, get a grounded chatbot 38
4In depth
Wire your apps together so repetitive work runs itself — no code. n8n and KNIME you can self-host and own; Make and Zapier are hosted drag-and-drop builders; Dify focuses on LLM-powered apps. → Pick by what you want: the fastest way to wire up popular apps → Zapier; visual multi-step flows with branching → Make; self-host it and drop into code when the nodes run out → n8n; clean, join and analyse data tables → KNIME; a chatbot over your own documents → Dify.
5Lessons 5
5.1 Create a chatbot that answers questions from your PDF
Dify is an open‑source visual studio for building production‑ready LLM apps such as RAG chatbots, agents and multi‑step workflows without writing code.
Chat with a paper from your own field in under 15 minutes
Try
What is the main finding of this paper?Paste the text into the message box on the public chat URL you received after publishing (the chatbot UI). Send the query and verify that the reply includes a quoted excerpt from your PDF with a source citation.
- Open the Dify Studio
- Click Create App → Chatbot and give it a name
- Open the Knowledge tab and upload a PDF
- Hit Publish to obtain a public chat URL
- Ask the deployed chatbot a question such as “What is the main finding of this paper?”
- You'll see Your published chatbot answered a question about your PDF with a grounded citation drawn from the document you uploaded
- Takeaway Dify turns any PDF into a live chatbot that answers queries instantly, pulling exact citations straight from the uploaded document
- Check Which tab do you upload the PDF in, and what does Publish give you at the end?
5.2 Create a searchable knowledge base from multiple documents
A Dify knowledge base aggregates PDFs, documents and web pages into a single searchable store that apps can query.
Do this first Create a chatbot that answers questions from your PDF
Turn a stack of sources into one searchable knowledge base
Try
In Dify Studio open the Knowledge section, create a new knowledge base named "AI Survey", then add sources: upload the files survey_part1.pdf and survey_part2.pdf, and add the web page https://openai.com/research. Save the base, attach it to the chatbot called "SurveyBot", and ask: What do these sources agree and disagree on?Paste this into the Knowledge tab of Dify Studio, using the Create Knowledge Base button then the Add Sources dialog to upload the PDFs and enter the URL. After linking in the chatbot settings, watch the document count rise as each source is chunked and embedded.
- Open the Knowledge panel and click Create knowledge base
- Click Add source to upload PDFs and enter a web page URL
- Wait for Dify to chunk and embed each source; observe the document count increase
- In the chatbot settings, select Attach knowledge base and choose the new base
- Send a multi‑source query such as “What do these sources agree and disagree on?” to the chatbot
- You'll see A single knowledge base holds multiple sources, and a chatbot answers a question using more than one of them with citations
- Takeaway A Dify knowledge base merges PDFs, documents and web pages into a single searchable store, enabling a chatbot to answer queries using information from multiple sources
- Check What has to finish before a source is searchable, and what does a multi‑source answer carry with it?
5.3 Swap the model used by your app
Dify lets you swap the underlying LLM provider for an app without rebuilding the app itself.
Do this first Create a searchable knowledge base from multiple documents
Replace the model behind your chatbot without rebuilding it
Try
Open the chatbot app in Dify Studio, click the **model selector** dropdown, choose a different provider/model (e.g., switch from OpenAI gpt‑4o to Ollama phi‑3), confirm the selection, then press **Save**.Paste this instruction into the Dify Studio interface while your app is open. After saving, re‑run the original grounded query and watch for a different response style indicating the new model is active.
- Open the knowledge chatbot created in the previous lesson within Studio
- Locate the model selector for the app and note the current provider/model
- Select a different provider from the model selector, such as a local Ollama or LM Studio model, or another cloud provider
- Re‑ask the same grounded question you used before
- Compare the new answer with the previous one to see the effect of the model change
- You'll see The same app answers the same question with two different models while the knowledge base stays unchanged
- Takeaway You can change a model instantly while preserving all other settings
- Check What stays untouched when you change the provider in the model selector, and what changes?
5.4 Create a multi-step workflow on the canvas
The Workflow app type provides a visual drag‑and‑drop canvas where nodes such as Knowledge Retrieval, LLM and conditional logic can be linked into multi‑step pipelines.
Do this first Swap the model used by your app
Wire your first nodes on the visual canvas
Try
Add a Knowledge Retrieval node pointing to my existing knowledge base, connect its output to an LLM node, then attach an If/Else node that routes short questions to one branch and long questions to another, finally run the workflow with the test input: "Explain quantum computing in simple terms."Paste this text into the Workflow editor prompt field on the canvas, then click Run. Watch the trace highlight each node sequentially and verify an answer appears at the end of the pipeline.
- Open Create App → Workflow in the Studio
- Add a Knowledge Retrieval node on the canvas and select your knowledge base
- Insert an LLM node and connect the retrieval output to it
- Place an If/Else node, configure branches for long and short questions, and connect it appropriately
- Run the workflow with a test input and watch the run trace illuminate each node
- You'll see A multi‑node workflow runs end to end on the canvas, retrieving from your knowledge base and producing a grounded answer through an LLM node
- Takeaway Workflow canvases let you link retrieval, LLM and conditional nodes so a grounded answer is produced through a visual pipeline
- Check Which node splits the canvas into a long‑question branch and a short‑question branch, and what feeds the LLM node?
5.5 Create an agent that calls tools
The Agent app type lets an LLM decide when to call tools or retrieve knowledge to achieve a goal, combining reasoning with external actions.
Do this first Create a multi-step workflow on the canvas
Give your app the ability to act by combining a knowledge base with external tool calls
Try
Find the relevant finding in my documents and check it against an external source.Paste this text into the Goal Prompt box when configuring your Agent app in the Dify Studio. After running, watch the Reasoning Trace panel to see when the agent reads the knowledge base versus calls the enabled tool.
- Open Create App → Agent in the Studio
- Attach your knowledge base to the new agent
- Enable at least one Tool for the agent
- Enter a goal‑shaped prompt such as
Find the relevant finding in my documents and check it against an external source. - Run the agent and observe its reasoning trace
- You'll see The agent completes its goal and the reasoning trace shows both knowledge‑base retrievals and tool invocations
- Takeaway Agents decide when to fetch stored information and when to invoke external tools to achieve a goal
- Check What does the agent's reasoning trace show that a plain chatbot's would not?
6You’ll know it worked 183 checkable outcomes in this chapter
- ✓Lab members receive a Slack message each Monday with 10 summarized abstracts
- ✓Employees receive a chatbot reply that includes the exact policy clause number and the ticket volume for common HR questions drops within weeks
- ✓Sales rep receives a Slack message with an edit link to the new PandaDoc proposal
- ✓Drafted outreach appears in Airtable with status 'Awaiting review' for each candidate
- ✓Only the configured percentage of traces are sent to the collector
- ✓All remaining rows have a sample size above the chosen cutoff (e.g., n≥10)
- ✓Submit the test form and see the data appear as the node's output in the execution log
- ✓Submitting a test form with different budget amounts sends the appropriate email and logs to the correct Google Sheet row
183 outcomes in all — one per recipe below.
7FAQ, Tips & How-to 285
one problem, one solution, one action
Research & data tools7
Manual Excel cleanup of plate‑reader data takes forever
A 45-minute manual Excel clean-up is replaced by a reproducible, one-click pipeline any lab member can run.
Need to add functional info to many FASTA entries
Hundreds of sequences are annotated overnight without writing a single line of Python.
Need a weekly QC report that flags visit values
Every data transfer is quality-checked consistently and a report is ready before the weekly data review call.
DifySummarizing research papers in a spreadsheet takes days
Early-stage literature synthesis that takes days in a spreadsheet is turned into an interactive Q&A session.
Engineers have to scan dozens of nightly CSVs
Engineers arrive in the morning with a clear overnight pass/fail report instead of manually scanning dozens of CSV files, and out-of-spec runs are visible before the morning stand-up.
Want a weekly list of new specialty research
The physician gets a curated weekly reading list in their inbox instead of relying on catching relevant new studies between conferences.
Knowledge & docs4
Need a constantly updated Notion paper list
The team's Notion library is updated daily with relevant preprints so no one misses a key paper.
New lab members can’t locate SOPs quickly
New lab members find protocol answers in seconds instead of hunting through shared drives or asking a senior colleague.
Employees stuck waiting for policy answers by email
Finance handles fewer one-off policy queries by email; employees get instant, policy-grounded answers any time of day.
Need instant competitive talking points on a call
Reps handle competitive objections confidently without putting prospects on hold to find the right slide deck.
Internal tools & ops13
Jumbled Gmail inbox
The inbox is pre-sorted before a human reads it, cutting triage time by roughly half.
Labeled Gmail attachments need to go into a spreadsheet
Incoming data files land in the spreadsheet automatically instead of requiring manual download and paste.
Employees waiting on policy answers
HR ticket volume for common policy questions drops, and answers are traceable to the exact policy clause.
Getting a generic onboarding list
Every new hire gets a checklist tailored to their actual role, not a one-size-fits-all template — logic that would need a separate branching app in a pure no-code tool is just a few lines of code inside the workflow.
Getting purchase approvals stuck in email threads
Purchase approvals that required email chains and manual ERP entry are handled in a single Slack thread, with a full audit trail.
Employee leaving
No departed employee retains access beyond their last day; every offboarding step is tracked and timestamped in one Slack thread.
Never miss a birthday or anniversary
Milestone recognition happens consistently for every employee without anyone monitoring a calendar or remembering dates.
Can't find the right leave form after asking HR
HR is freed from repeat policy queries by email; employees get instant answers at any hour and are guided to the right form for transactional requests.
Tagging a Zendesk ticket as a bug
Support and engineering stay in sync without manual copy-paste; every filed bug has a Linear issue and agents can tell customers the issue number.
Unsure if a part’s datasheet was updated
The hardware team is alerted to supplier datasheet changes the morning after they go live, before a revised component reaches the production line unnoticed.
Field‑complaint emails sit unread
Every field complaint is captured in the CAPA system within minutes of receipt and arrives with an AI-drafted triage summary, reducing the time from complaint receipt to initial risk classification and keeping the complaint-handling timeline compliant with ISO 13485 requirements.
Lab results stuck in a shared inbox
Results reach the right clinician's queue within minutes of arriving instead of sitting in a shared inbox, with a full audit trail of when each result was filed and viewed.
Referral letters stuck in a fax tray
Referral letters that used to sit in a fax tray are visible in a shared queue within minutes, with routine intake sorting drafted automatically and every triage placement confirmed by a human.
Content & marketing5
Tagging a new GitHub release
A polished changelog entry is live within a minute of tagging a release — without a human writing it.
Publish one CMS post to LinkedIn, X, Facebook and email
A single publish action distributes to four channels in under a minute, keeping brand voice consistent.
New sign‑up forms don’t trigger welcome emails
New signups enter the nurture flow within seconds, improving first-impression timing.
Blank page for a new piece
Writers start every piece with a consistent, on-brand brief instead of a blank page, cutting brief-writing time by around 80%.
Posting a job on three sites is slow
A role that took 30 minutes to post manually across three platforms is distributed in under a minute from a single Airtable record.
Forms, surveys & feedback7
Stop getting noisy form alerts
The team is only interrupted for responses worth acting on, while every submission still lands in a searchable Airtable base — built entirely by connecting modules on a canvas.
Survey responses that are mixed up or incomplete
Data cleaning that usually takes a day in SPSS or R is a reusable, auditable KNIME workflow.
Job applications pile up
Recruiters only review candidates who passed the baseline screen; rejection emails go out within minutes rather than weeks.
Managers have to approve PTO manually
Leave approvals that required chasing managers by Slack and manually updating a spreadsheet are handled end-to-end in one automated loop.
HR has to chase employees for reviews
HR no longer chases individuals manually; submission rates improve because reminders are timely and personalised.
Need timely CSAT after tickets close
CSAT collection is automatic and timely; every low score is visible to the team lead within minutes of being submitted.
CV emails arrive as PDFs
Consultants open the ATS to find a pre-structured record instead of a raw PDF, cutting the time to first review and reducing manual data entry per application.
Commerce & payments7
One‑time sales and subscriptions land in the correct QuickBooks records
One-time sales and recurring subscriptions land in the correct QuickBooks bucket automatically, and customers get a receipt — no manual bookkeeping split at month-end.
Want automatic review requests
Review volume increases without any manual follow-up effort from the shop owner.
Invoices stuck in email for days
Invoice approval cycles that took days of email back-and-forth are resolved in a single Slack thread, with the accounting entry created automatically.
Expense reports need receipt chase and manual entry
Expense reimbursements that required chasing receipts and manual data entry are handled end-to-end in one automated flow.
Overdue Xero invoices aren’t being chased
Days-sales-outstanding drops because every overdue invoice is chased on schedule, with no one manually monitoring the aged-receivables list.
Bank transactions don’t line up with the ledger daily
Month-end close takes hours instead of days because routine transaction matching is done automatically each morning.
Failed card payments go unnoticed
Involuntary churn from failed card payments is reduced because recovery emails go out automatically within minutes, not days.
CRM & sales10
Demo bookings fall through
Every booked demo is automatically tracked and followed up — no lead falls through after the first call.
Closed-won deals aren’t announced or tracked
Wins are celebrated publicly and revenue is logged automatically — no hunting in the CRM for month-end numbers.
New leads are a mess
Sales reps only receive pre-qualified leads and know exactly which contacts to prioritise, cutting time-to-first-contact on hot leads.
Deal reaches proposal stage and reps avoid copy‑paste
Reps skip the 20-minute copy-paste from CRM to proposal tool; the draft is ready to personalise in seconds after staging a deal.
Signature received but deal stays open
The gap between signature and CRM update disappears, and customer onboarding starts the same minute the contract is signed.
Can’t tell if a new lead fits your target customer
Reps know within seconds whether a new lead fits the ICP and have the company context they need before picking up the phone.
Missed demo not followed up
No-shows are re-engaged automatically within the hour while the missed meeting is still fresh, recovering deals that would otherwise fall silent.
Large‑deal quote stuck in email chains
Large-deal discounting is governed without a cumbersome email chain; reps get a same-hour decision on whether to send.
Pitch email lands in your inbox
Every inbound pitch is captured and visible to the team within minutes, with key fields pre-filled so analysts skip the copy-paste and go straight to first-pass diligence.
Need a personal outreach draft for each candidate
Consultants start from a personalised first draft rather than a blank page, cutting message prep time per candidate while keeping a human in the loop before any contact is made.
Dashboards & analytics7
Tired of opening five dashboards every morning
A single morning email replaces manual tab-switching across five dashboards.
Get batch predictions from CSV without coding
A working predictive model is built and validated without writing any code, and retraining is a one-click re-run.
Can’t keep revenue up to date
MRR and churn are always up to date without anyone manually pulling Stripe reports, giving finance a live revenue view.
When monthly spend drifts over budget
Finance stops sending bulk monthly reports that no one reads; only the managers with a real variance see an alert, and they see it on day one.
Need fast boardroom‑ready spend analysis
Finance produces a boardroom-ready spend analysis in one click instead of spending an afternoon in Excel pivot tables.
Monthly headcount and attrition reporting takes half a day
The monthly people-metrics pack that previously took a half-day in Excel is produced in one workflow run, consistently and without formula errors.
Need a single KPI sheet for all founders
The whole portfolio is summarised in one spreadsheet view before the Monday LP call, with no analyst manually reading and transcribing twelve founder emails.
Booking & scheduling3
New Calendly booking happens
The whole team sees new bookings in real time so the right person can prepare.
Changing a candidate’s status to schedule interview
A status field change replaces three manual emails and calendar invites, cutting interview-scheduling time from hours to minutes.
Patients miss appointments
The no-show rate drops because every patient gets a same-day nudge, and the front desk no longer has to manually call each no-show to rebook.
Customer & client portals10
Routine FAQ traffic is deflected automatically, cutting first-response time from hours to seconds.
Tickets go silent past their SLA response time
SLA commitments are enforced automatically; no critical ticket sits silent because someone missed a queue check.
Customer shows frustration in a ticket
At-risk customers get a senior response within minutes instead of hours, before the situation escalates further.
Support messages from email, chat and social
Support messages from three channels land in one queue — agents work one tool instead of switching between tabs and missing messages.
Blank reply box on new tickets
Agents start from a relevant draft instead of a blank reply box, cutting average first-response time significantly on common issue types.
VIP tickets get lost in the queue
VIP customers never wait in the standard queue; account managers are looped in from the first message so they can manage the relationship proactively.
Agents stuck searching docs while handling tickets
Average handle time drops because agents find answers in seconds rather than searching the knowledge base or pinging a colleague.
Support emails land in the wrong queue
Support emails land in the right queue within seconds, cutting first-response time and preventing high-priority issues from sitting unseen.
Agents need to hunt for help articles
Agents resolve common issues faster because relevant articles are surfaced automatically instead of searched manually.
Users ask routine support questions
Routine tier-1 questions are deflected without a human agent, reducing ticket volume while still capturing contacts who need real help.
Want to pick blog topics yourself
Trigger the workflow with a slash command and fill a form for blog details
Need a key that can only read my automations
Securely connect n8n to Claude by creating an API key that can read workflows
Use n8n's execution history to pinpoint where a workflow fails
Want to trigger Outlook, Teams or Word from a workflow
Use the Agent 365 trigger node in n8n to securely invoke Microsoft 365 APIs from a workflow
Webhook from Agent 365 won’t reach Teams or Outlook
Link the webhook endpoint from the Agent 365 blueprint with an LLM chat model node and optional memory or external tool nodes to build an AI teammate
Non‑technical team members can’t handle API keys
Users can connect to services like Gemini or Slack without handling API keys
Need quick Slack updates or transcript conversion
Users can start from a ready-made workflow and modify placeholders
A focused 28-48 hour event surfaces high-value automations
After initial build, a stakeholder maintains and expands the workflow
Curriculum checks that take hours
Automate curriculum checks against n8n standards, reducing processing from six hours to ten minutes with a 95 % pass-rate
MCP auto-save feature — ensures workflow persistence
Users can rely on n8n to automatically save their work, preventing data loss
Instance-level MCP one-click connect — cross-platform integration
Users can instantly link their n8n instance to external platforms with a single click
Ask AI Assistant workflow — answers docs queries with low error
n8n can host a self-built AI assistant that reliably answers documentation questions
Podcast generation workflow — auto-create 30-minute episodes
n8n can orchestrate complex media production tasks, such as generating podcasts on the fly
AI intent to workflow generation — auto-build from user intent
n8n aims to let users specify what they want and have the platform generate the necessary workflow automatically
Community-driven feature development — rapid two-month cycle
Listening to user feedback can accelerate feature delivery from idea to production
Check a Gmail label regularly
Automate email checks by polling Gmail at set intervals
Raw email with attachments
Turn raw MIME into a format that can be parsed for URLs
Swap email providers without changing core logic
Need more threat‑intel sources
Enhance detection by integrating more APIs
Want to know if a link is safe
Leverage VirusTotal to get a quick maliciousness score for URLs
Not sure if a detection is real phishing
Automate labeling based on a criticality rating from VirusTotal
When my Ubiquiti firewall blocks traffic
Receive real-time block alerts from a Ubiquiti router into n8n
Blocked URLs lack visual context
Add visual context to blocked URLs with URLScan data
Want one threat report from multiple scanners
Generate a single, actionable threat report in seconds
Router threat logs need automation
Use the router's built-in threat management logs to start workflows
Need to copy and adapt a workflow quickly
Build workflows that can be copied and adapted easily
Want every Gmail email to fire the flow and disappear
Tailor the email workflow to process all messages or archive automatically
New incident arrives
Automatically launch the AI-augmented triage pipeline whenever a ticket arrives
Node positions drift while working
Keep canvases tidy by locking node group positions
Trigger actions on your PC from a workflow
Automate local tasks like opening browsers and scraping data via workflows
Don’t know which user filled out a form
Authenticate users via Google OAuth before processing form data
Get instant debugging insights directly in the workflow editor
Prototype UI concepts and product experiments without immediate release pressure
Need a private workflow tool to talk to MCP using tokens
Connect a private n8n instance to the MCP endpoint using OAuth for token-based authentication
Use the visual editor to see exactly which node caused a failure and edit it directly
Can't turn prompts into code
Learn how to activate MCP in n8n and link it to an LLM like Claude
Support tickets submitted on a form
Automate ticket handling from Jotform to your SaaS workflow without custom code
Want tracing without touching code
Turn on OpenTelemetry tracing in n8n without code changes
See end-to-end latency by viewing workflow as parent and nodes as children
Want to attach extra info to a workflow step
Add custom data to a node's span that appears in the waterfall chart
Traces flooding your logs
Control how many traces are exported to reduce overhead
Want only high‑level workflow visibility
Reduce detail to workflow-level visibility when needed
Downstream services can’t see your trace ID
Allow downstream services to create child spans automatically
Want to try distributed tracing quickly
Set up a tracing infrastructure quickly for experimentation
n8n's OTEL implementation works with existing tracing backends
OpenTelemetry is available out of the box in n8n 2.22
You can access the full list of chapters and cards for the n8n AI course
Knowing the total number of chapters helps you gauge the course length and structure
Every lesson card follows a consistent layout, making navigation predictable
Understanding the three moves lets you follow the course flow and know what to expect in each session
Need a clear learning‑goal paragraph
Articulating a concrete automation target gives you a personal project to apply the course material
Want to share a paragraph with the whole class
Sharing your problem statement gets feedback and anchors your learning in a real use case
You can see how the pre-built AI agent behaves before modifying it
Clicking a node reveals its configuration, helping you learn what each piece does without memorising settings
If the agent remembers prior conversation, you can build multi-step interactions without restating context
A correctly set constraint makes the agent politely decline out-of-scope queries
Need the agent’s reply shown in your Mattermost chat
The workflow automatically posts the agent's reply into the same Mattermost channel
Want to try out a workflow change without breaking the live version
Copying the shared workflow lets you experiment without affecting the original bot
Need the bot to talk in a specific style
Changing the system prompt rewrites how the bot talks, letting you add constraints or style
The agent remembers prior turns only in the same Mattermost channel, not across channels
Block messages containing a forbidden word
You can prevent certain inputs from reaching the agent by adding a conditional check
Don’t know what language a question is in
Detecting the input language and passing it to the agent improves handling of non-English questions
Want all bot replies to finish the same way
Appending a structured footer to every answer gives users consistent guidance
Want a ready‑made document Q&A setup
You can quickly set up a document Q&A system by importing a ready-made n8n workflow
When a question cannot be answered from the document, the AI will refuse or say it's not in the source
The same QA flow works for any reachable URL, letting you reuse the pattern for your own research
Changing the slice size in the Prepare Context node lets you see how more or fewer characters affect answer quality
Different source formats (e.g., Wikipedia vs. PubMed) affect how well the model can answer, revealing strengths and limits of the fetch-and-stuff method
Need a short summary of any text
Typing a message that starts with "Summarize this:" routes the text to the summarisation agent
Need to pull key facts from a message
Messages that start with "Extract key facts from:" are routed to the extraction agent
Route messages by keyword
A Code node can examine the user message and set a route variable based on keywords, ensuring predictable routing
Split flow based on code decision
An IF node reads the `route` value from the Code node and directs execution to the appropriate AI branch
Need an extra translation step
You can augment the workflow with another keyword check, an extra IF branch, and a new translation agent
Ask a research question in plain English
You can ask a research question in plain English and receive AI-summarised abstracts from PubMed
Want to pull PMID list from PubMed search results
You can capture the list of PMIDs from the first API call to feed subsequent requests
Need citation counts for PubMed papers
You can enrich PubMed results with citation counts by calling the Semantic Scholar API for each PMID
Need citation count and PMID from an API response
A Set node lets you rename and store the fields you need for later use
Need citation numbers in AI prompt
Modifying the Prepare Context Set node to add citation data ensures the LLM mentions it in its summary
Separate PubMed IDs from plain text
You can separate valid PMID inputs from free-text queries in one step
PubMed request returns 404
Unexpected failures can be handled without breaking the whole workflow
The visual log shows exactly which nodes ran and whether they succeeded or errored
Want to label each workflow branch with its outcome
You can tag each execution branch with a clear outcome for later aggregation
Want one node to receive data no matter which path runs
A single downstream node can receive data regardless of which branch fired
Need to explain workflow paths
Adding notes keeps future maintainers aware of each path's purpose
I have a research question
Sending a natural-language query starts the chain of nodes that fetches papers and synthesises an answer
Need a single keyword for a follow‑up PubMed search
An AI Agent can parse the first synthesis and output only the most important keyword for a second PubMed query
Want a PubMed search link from a keyword
A Code node can programmatically create the correct eSearch endpoint using the keyword
I need additional PubMed articles
Repeating the PubMed nodes with a new URL pulls additional papers that complement the first set
Combine original and second‑round papers
A Set node can merge two arrays of paper objects into one collection for final synthesis
A long list of papers from two rounds
Feeding the combined paper list to an AI Agent lets it produce a cohesive research landscape overview
Want a quick literature overview from a research question
You can automatically turn a research question into a formatted Markdown table of papers with key details
You can debug the AI extraction step by running it alone on a single abstract
Need citation numbers in my paper list
You can extend the literature pipeline to fetch citation counts from Semantic Scholar and display them in the final table
Paper missing citation data
When a paper has no citation data yet, the pipeline should still produce a table entry without breaking
Lost a saved workflow and want it back
You can quickly restore the demo workflow without rebuilding it
Need the workflow to look for just one research question
Tailoring the search term focuses the pipeline on papers you actually need
Pull key details from research papers
Running the workflow produces a table with method, sample size, key finding, and limitation for each paper
Getting rid of papers without domain keywords
Filtering out papers whose methods lack domain-specific keywords reduces irrelevant rows
Tiny studies polluting results
Dropping tiny studies prevents noisy data from contaminating the summary
Missing a data point in AI extraction
Adding a new extraction target (e.g., statistical test) enriches the structured output
Need to create an app fast without coding
You can access Dify's drag-and-drop interface to start building apps
Need a brand‑new chatbot project
Creates a fresh app of type Chatbot that you can configure and publish
Make your chatbot reachable online
Makes the chatbot reachable via a URL or embeddable widget
Want the chatbot to sound different
Changing the chatbot's instruction text reshapes its output without affecting grounding
You can instantly see which LLM your app is using without opening any code
Running the same query on two models lets you see concrete differences in output quality and style
Testing both models with identical input provides concrete data to decide which fits your constraints
Unsure which LLM to pick for new chats
Designating a winner makes future chats automatically use the chosen model without extra steps
Need a way to build multi‑step workflows
You can begin building multi-step workflows by creating a Workflow app in Dify's Studio
Can’t find the right info in your knowledge base
Linking a Knowledge Retrieval node lets the workflow pull relevant text from your knowledge base
Need separate paths for long and short questions
An If/Else node enables conditional routing so different question lengths follow separate paths
Factual and open‑ended questions
A classifier node can split incoming queries into categories so each follows a tailored processing path
Question types need separate workflow branches
Linking each classification outcome to a different node chain lets the workflow handle distinct question types appropriately
Running one example of each class proves the classifier and branching logic work as intended
My bot can’t see my files
Linking a knowledge base lets the agent retrieve information from your documents while reasoning
Enumerating the exact nodes or tools clarifies implementation requirements and avoids missing functionality
Stating whether the solution is a public chat URL, embedded widget, or Service API determines integration and access strategy
Need a tidy spot for Docker Compose files
Creating a dedicated folder keeps the Docker Compose configuration and related files tidy and isolated
Need current weather data
Retrieves external data that can be used later in the workflow
Need to send data with POST to OpenRouter’s chat API
Setting POST and the correct endpoint directs the call to OpenRouter's chat API
Need to hide your API key in a request
Providing a Bearer token authenticates your request without exposing the key in the workflow
Need to pick a model and set your question for OpenRouter
The request payload tells OpenRouter which model to run and what message to answer
Running an API call in a workflow
Running the configured HTTP Request returns a free-model response directly in the node output
Can’t pass a generated value into my prompt
A preceding node can generate a value that becomes part of the HTTP request payload
Need to use the AI’s reply in another step
You can route `choices[0].message.content` to another node for further processing, such as writing to a file or sending a message
Running the workflow with multiple inputs confirms that dynamic prompting and response handling are robust
What is Dify and what can I use it for?
Dify is an open-source platform for building AI-powered applications — like chatbots and question-answering assistants — without writing code. You design your app visually, connect an AI model (such as ChatGPT or Claude), and optionally upload documents for the AI to answer questions from. It is especially well-suited for non-technical users who want a working chatbot over their own documents in a short time.
Should I use Dify Cloud or self-host Dify — what is the difference?
Dify Cloud means signing up at dify.ai and using it in your browser with no installation — the easiest option for beginners. Self-hosting means running Dify on your own computer or server with Docker, which gives full data privacy and no usage limits beyond your hardware. Cloud is best for building quickly; self-hosting makes sense if your data is sensitive (e.g. unpublished research) or you need more documents and storage than the free cloud tier provides.
How does Dify compare to n8n — which should I use?
Dify is built specifically for AI chatbots and knowledge-base apps, with RAG, prompt engineering, and chat interfaces built in — you can have a working chatbot in under an hour with no coding. n8n is a general workflow-automation tool with hundreds of integrations that treats AI as one optional component; it has a steeper learning curve. For a chatbot over your papers, Dify is the right choice; n8n shines when you need to connect many external systems alongside AI.
What is a Dify Knowledge Base and how does RAG work?
A Knowledge Base is a collection of your own documents (papers, notes, PDFs) that you upload into Dify. When someone asks a question, Dify first searches those documents for the most relevant passages, then sends those passages along with the question to the AI model — a technique called Retrieval-Augmented Generation (RAG). The result is an AI that answers based on your specific materials rather than just its general training, making it far more accurate for specialized topics.
What file types can I upload to a Dify Knowledge Base?
Dify supports a wide range of formats: TXT, Markdown, PDF, HTML, Excel, Word, CSV, PowerPoint, and more, up to a per-file size limit. For most uploads of paper PDFs or Word notes, the standard mode works without extra setup; an alternative processing option unlocks the broader format list including presentations.
How do I upload my papers and notes to a Dify Knowledge Base step by step?
Go to the Knowledge section, click Create Knowledge, and name it. Drag-and-drop or browse to select your files, then choose automatic text chunking (recommended for beginners) and a high-quality indexing mode for better accuracy. Click Save and Process — indexing may take a minute or two. Once complete, attach this Knowledge Base to your chatbot app, and the bot can answer questions based on its contents.
What AI models does Dify support — can I use ChatGPT, Claude, Gemini, or local models?
All of them. Dify supports OpenAI (GPT), Anthropic (Claude), Google (Gemini), and many more cloud providers, and it also supports local models running on your own computer via Ollama — meaning you can run open-source models for free with no per-message cost, though that needs a capable machine. You add any provider under Settings → Model Providers by pasting in your API key.
Need others to try your bot without logging in
Anyone can access the bot without a Dify account via a shared URL
Need a local AI assistant server
Learn how to install and start Dify on your own machine using Docker
Need Notion pages in a knowledge base
Learn how to bring Notion content into a Dify Knowledge Base and keep it updated
Want a fast literature search without manual digging
Get a ranked reading list in minutes instead of hours of manual searching
Need to hand off a workflow but keep credentials private
Learn how to send a workflow to someone else without sharing credentials
Get a quick introduction to n8n's interface and basic concepts
What is n8n and what can I use it for?
n8n is a visual workflow automation tool that connects different apps and services so they can pass information between each other automatically, without you writing code. You build automations by placing 'nodes' — each representing one app or action — on a canvas and drawing connections between them. A researcher could, for example, set up n8n to automatically collect papers from PubMed, summarize them with AI, and write the results to a Google Sheet — all triggered on a schedule.
Can I run a workflow on a schedule — for example, every day at 9 AM?
Yes. Use the Schedule Trigger node as the first node in your workflow. You can choose simple intervals (every X minutes, hours, or days) without any technical knowledge, or use a cron expression for precise schedules like 'every weekday at 9 AM'. The node also has a timezone setting so your schedule reflects your local time rather than the server's timezone.
How do I connect n8n to an app like Google Sheets or Gmail?
Each app integration in n8n requires a 'credential' — the login or access key that lets n8n talk to that service on your behalf. When you add a node for a service, n8n prompts you to add a credential. For Google services you click 'Connect my account' and log in via OAuth (the familiar 'Sign in with Google' popup). For other services you paste in an API key copied from that service's settings page. Credentials are stored securely and reused across all your workflows.
Are there ready-made workflows I can start from instead of building from scratch?
Yes — n8n has a public template library with thousands of community-built workflows covering a huge range of use cases. You can browse by category, preview what each workflow does, and load one directly into your editor with a single click. Templates are an excellent way to learn how experienced users structure their automations and a much faster starting point than building from scratch.
Do I need to know how to code to use n8n?
No — most workflows can be built entirely by clicking and dragging without writing a single line of code. However, n8n does have a steeper learning curve than some alternatives like Zapier: features such as expressions, error handling, and connecting AI tools require some patience and self-study. Non-coders who are willing to practice consistently report becoming comfortable with the interface after building a few simple workflows.
What is a node in n8n?
A node is a single building block in your workflow — it represents one action, service, or piece of logic. For example, a 'Gmail' node can send an email, a 'Google Sheets' node can write a row of data, and a 'Code' node lets advanced users add custom logic. Trigger nodes are a special type that start the whole workflow when a specific event happens (like a new file being uploaded or a schedule being reached).
How does n8n compare to Zapier or Make — which is easier for a complete beginner?
For someone with no technical background, Zapier is the easiest of the three: it uses a step-by-step wizard and requires no technical decisions. Make sits in the middle with a visual canvas. n8n has the steepest learning curve because it uses expressions, webhooks, and a more flexible but complex interface. That said, n8n is the most powerful and cheapest at scale, and its growing library of AI tools and templates is closing the gap.
What is a webhook and why does n8n use it?
A webhook is like a doorbell: instead of your workflow constantly checking whether something new has happened, an external app rings the bell (sends a message) the instant an event occurs. In n8n, the Webhook node gives your workflow a unique URL address; when another service sends data to that address, your workflow starts immediately. This is more efficient than scheduled polling and enables real-time responses — for example, triggering a workflow the moment someone submits a form.
How do I get my workflow to actually run automatically — it only works when I click 'Test'?
You need to activate the workflow. In the top-right corner of the editor there is an 'Inactive / Active' toggle switch. While it is set to Inactive, the workflow will not respond to real triggers — it only runs when you manually click Test. Switch the toggle to Active (it turns orange) and save, and your workflow will begin listening for events and running on its own.
Want a task to start every night at 7 PM
The Schedule node lets you define when a workflow should start, using cron expressions or simple interval settings. It’s the entry point for any recurring automation.
Get the latest tech news
The HTTP Request node can call any REST endpoint. By supplying your Perplexity API key in the headers, you retrieve JSON‑formatted news data for further processing.
Need LinkedIn post ideas from a news summary
The Gemini node (Google AI) performs text completion. Feeding it the news summary lets you generate multiple stylistic LinkedIn post variations automatically.
Want to convert a base64 image into a Drive file you can share
The Google Drive node can upload binary data as a file, then return a public web view URL. Converting the image from base64 to binary first ensures proper upload.
Collect an email and preferred install date via a web form
A Form Trigger node creates a public URL that serves a web form. When the form is submitted, it starts the workflow and outputs the entered fields as JSON data.
Tired of re‑submitting a form for each test
Pinning data on a trigger node stores a fixed JSON payload, so each execution returns that same data without needing to re‑submit the form. This speeds up iteration and debugging.
Split requests that occur within a week from later ones
The IF node evaluates expressions per incoming item. By comparing the preferred install date to ‘now + 7 days’, you can split the flow into true (within a week) and false branches.
Need to ping Slack with requester email and install date
The Slack “Send Message” action uses credentials to post into a workspace. You can compose the message by mixing static text and expressions that pull values from previous nodes.
Workflow stuck in test mode
Activating a workflow switches it from test mode to production. Once active, every incoming request (e.g., via the form URL) triggers real executions that are logged under the Executions tab.
A user submits your web form
A Form Trigger captures data when a user submits a web form and passes that data as the first node in an n8n workflow. It works by listening to a specific form URL, so any new submission automatically fires the workflow.
Web form submissions need to be recorded
The Google Sheets node can append or update rows using data from previous nodes. By mapping form fields to sheet columns, each new lead is recorded automatically.
Want to capture when a form is submitted
n8n expressions let you compute values on‑the‑fly. Using double curly braces with $now() inserts the current date/time, which can be stored alongside form data.
Leads that can’t afford the minimum spend
The If node evaluates a condition and routes execution down “true” or “false” branches. By checking the budget field, you can automatically flag leads that don’t meet your minimum spend.
Leads with low budgets are ignored
A Filter node passes data only when a condition is met. It’s useful for silently discarding leads that fall below a threshold without extra branching.
Leads have varying budgets
The Switch node evaluates a value against multiple cases and directs execution down different paths. It lets you send high‑budget leads one email template and low‑budget leads another.
Add a source tag to each row before saving
A Set node creates or overwrites data fields using static values or expressions, allowing you to enrich payloads (e.g., adding a “source” tag).
Execution Log — debug workflow runs
The Executions view shows each step’s input and output, helping you verify data mapping and spot errors. It’s essential for troubleshooting new workflows.
Want a workflow to kick off on a form fill‑out
The 'On Form Submission' node captures data from a custom form you define within n8n, providing the initial payload for the rest of the workflow. It works by exposing a temporary URL that can be embedded in any web page or shared directly.
Form submissions need a permanent record
The ‘Google Sheets – Append/Update Row’ node writes incoming data into a spreadsheet. By mapping each field to a column you create a persistent record without writing any code.
When student entries slip into the flow
A Filter node evaluates a condition and only passes data downstream when the condition is true. It’s useful for branching logic without extra code.
Need to handle different jobs separately
The Switch node creates multiple branches based on different condition values, allowing parallel handling of distinct cases (e.g., engineer vs doctor).
Want to alert each lead automatically
The Gmail ‘Send Message’ node sends an email using your connected Gmail account. By inserting variables from previous nodes, you can personalize the message for each lead.
Engineer and doctor email paths need syncing
A Merge node with mode ‘Wait for All’ synchronizes multiple incoming paths, allowing you to continue processing after all conditional routes have finished.
Need a chatbot that gives only ranked lists
By attaching an OpenAI chat model to the AI Agent node and configuring system prompts with specific output formatting, you can force the LLM to return only the desired list items, reducing token usage.
Chatbot forgets previous prompts
Adding Simple Memory to an AI Agent stores the last N interactions, allowing follow‑up questions to inherit prior context without re‑prompting the user.
User fills out a sign‑up form
The "On Form Submission" trigger lets you build a web form directly in n8n, map fields to variables, and output structured JSON for downstream nodes.
New user signs up on my form
Connecting a Google Sheets "Append Row" node to the form trigger writes each submission as a new spreadsheet row, enabling persistent storage without code.
LLM spits out random text instead of a neat ranked list
Including one or two formatted examples inside the system prompt (one‑shot) guides the model to mimic the desired output style, striking a balance between zero‑shot and few‑shot prompting.
Unsure which branch runs first in a multi‑branch workflow
n8n runs each branch sequentially based on canvas position: top‑most to bottom‑most, left‑most when heights match. Understanding this lets you design workflows without manually reordering execution.
Need extra actions in my workflow without coding
Community nodes are third‑party packages that add new actions to n8n without writing code. Installing them from the npm registry lets you use pre‑built integrations like Amplify or MCP directly in your workflows.
Want to handle each automation platform separately
A Manual Trigger lets you start a workflow on demand, while the SplitInBatches node can turn an input array into separate items for downstream processing.
Want to keep workflow info without extra spreadsheets
n8n Data Tables act like built‑in spreadsheets, allowing you to read/write rows directly from workflows, eliminating extra API calls to Google Sheets or Airtable.
Need an answer from a workflow via chat
The Chat Hub provides an internal LLM interface that can invoke n8n workflows through a chat trigger, enabling conversational automation.
When I add or update a row in my spreadsheet
A Google Sheets trigger node watches a spreadsheet for added or updated rows and fires the workflow each time. It requires setting up OAuth credentials in Google Cloud, then selecting the sheet and event type.
Want to turn raw order data into a ready‑to‑send email
The OpenAI node can call a language model (e.g., GPT‑4o) to transform incoming JSON into a custom summary. By passing fields as variables, the prompt adapts to each order without code.
Every time a new order row appears, get an automatic email with the order summary
The Gmail node sends an email using the subject and body produced by the OpenAI node. Mapping the JSON fields directly avoids extra parsing steps.
Node Types Overview — building blocks of n8n workflows
n8n workflows consist of four core node categories: Trigger (starts execution), Action (performs a task), Data Transformation (modifies data), and Logic (controls flow). Knowing each type helps you design clear, maintainable automations.
Need a daily 6 AM start for your automation
The Schedule trigger node initiates any n8n workflow on a timed interval. By setting the interval to 'Days' and specifying hour/minute, you can have the workflow start automatically every morning at 6 AM.
Need today’s weather for a ZIP code
The Open Weather node is a pre‑built integration that calls the OpenWeather API. After adding your API key once, you can reuse the credentials in any workflow to retrieve live weather information.
Want to email the current temperature and location
The Gmail node sends emails using OAuth credentials. You can insert data from previous nodes by dragging fields into the subject or body, creating fully personalized messages.
Different budget amounts in form leads
A Form Submission trigger captures user input, and a Switch (logic) node can branch the flow using numeric comparisons. This pattern lets you automatically separate high‑value from low‑value leads.
Can't find a built‑in node for air quality
The HTTP Request node lets you perform GET or POST calls to any public endpoint. By supplying URL, method, headers and query parameters, you can integrate services that n8n doesn’t ship with pre‑built nodes for.
Need an external app to kick off your automation
A webhook node creates a public URL that can be called by any service to start a workflow. It acts as the entry point, similar to a door opening when someone knocks.
Need only certain values from incoming data
The Edit Fields node lets you map, rename, and filter properties of the incoming JSON so downstream nodes receive only what they need.
Want to keep collected form entries in a spreadsheet
The Google Sheets node uses OAuth credentials to write rows into a sheet, turning n8n into a lightweight database for collected form entries.
A user sends a Telegram command
The Telegram node can act as a trigger that fires whenever a user sends a message to your bot, enabling chat‑based automation.
Need a chatbot response in your workflow
The OpenAI node sends a prompt to GPT‑4 (or other models) and returns the generated text, allowing you to build conversational agents.
Bot forgets what you said before
By storing previous messages in an n8n “Set” or “Data Store” node, you can feed past context back into the OpenAI prompt, giving the bot memory.
Need a workflow to run on its own
Publishing a workflow activates its triggers and allows it to run without manual execution, turning your design into a live automation.
Want your workflow to start on its own at regular times
The Cron node lets you define time‑based schedules (e.g., every hour) so the workflow runs without external input.
API call fails and the whole flow stops
Enabling ‘Continue On Fail’ on a node prevents the entire workflow from stopping when that node encounters an error, allowing later nodes to handle fallback logic.
New emails stay unorganized
Uses n8n’s Gmail node to watch new emails, extracts keywords, and applies labels like accounting, personal, or meetings via the Gmail API. Works by matching subject/body patterns to predefined label rules.
A lead submits your website form
Combines a Webhook trigger (for a quote form), a Twilio node to place a call, and a short delay so the call reaches you within seconds. The workflow bridges form data to a phone call automatically.
Need to copy Google Maps listings to a sheet and email them
Uses an HTTP Request node to query Google Maps search results, parses JSON with a Function node, writes rows to Google Sheets, then loops through each row with an Email Send node for cold outreach. Automates lead gathering from maps.
Can't sort through LinkedIn SEO postings
Leverages a Web Scraping node (or HTTP Request + HTML Extract) to search LinkedIn for SEO roles in Canada, stores results in Google Sheets, then runs a Function node that rates each posting on criteria (salary, seniority, remote). Provides a ranked list for focused applications.
Have a receipt photo and need each item listed
Connects an HTTP Trigger (or file upload node) to a OpenAI node that runs a prompt asking ChatGPT to list each line item. The result is parsed and written into Google Sheets, turning a photo of a receipt into structured data.
Want to add meetings or fire off emails with a chat message
Combines an Email Trigger, a Schedule Trigger, and OpenAI nodes to interpret natural‑language commands (e.g., “schedule meeting with John tomorrow at 3pm”), then uses Google Calendar and Gmail nodes to create events or send emails. Provides a conversational interface for daily tasks.
Website visitors can’t book meetings
Uses n8n’s Webhook node as an endpoint for a front‑end chat widget. The webhook forwards user messages to OpenAI, which returns suggested meeting times; those are then passed to Google Calendar to create events and to Gmail to send confirmations.
Need an invoice from a chat message
Integrates a Webhook (receiving invoice details), an OpenAI node that formats the data into a PDF using a template, then uploads the PDF to Google Drive and emails it via Gmail. Automates end‑to‑end invoicing without manual paperwork.
When a row in my Airtable base changes
By using n8n's Airtable node as a trigger you can poll a base at a set interval (e.g., every minute) and fire the workflow whenever a row’s ‘Last Modified’ field changes. This creates real‑time notifications without writing code.
Turn Airtable record data into a ready‑to‑send email
The Google Gemini node can be used as a Large Language Model step that receives data from Airtable and returns structured JSON containing an email subject and body. Using the “Structured Output” option forces Gemini to output a predictable schema, making it easy to map into later nodes.
Need to send personalized emails from a workflow
n8n’s Gmail node can send emails using OAuth credentials. By inserting expressions that reference the Gemini output, you can dynamically fill the email subject and HTML/text body, achieving fully automated personalized messages.
Airtable updates never trigger actions
A workflow remains inactive until it is published; publishing activates the polling interval for triggers like Airtable. Once live, n8n will repeatedly check the source and execute downstream nodes automatically.
The same set on /recipes, filtered by tool and role.
8Videos 20
Learn Make.com in 2026 | Complete Beginner's Guide to Automation | Everything You Need to Know
The quick coffee-break intro. Watch to decide if Make fits before committing to the 45-minute full guide.
Learning Zapier in 2026: AI Automation for Beginners
The most current single-video overview. Watch to see how much of Zapier is now AI rather than plumbing.
Zapier for Beginners: Build AI Automations Without Coding! Complete tutorial!
The clean beginner on-ramp. Start here to learn the trigger-action model before the agents video.
KNIME Tutorial for Beginners EP 01 | No-Code Data Analytics Explained | Assignment On Click
The start of a structured beginner series. Watch if you want to learn KNIME step by step across short, focused episodes.
Get Started with Nodes & Workflows in KNIME
Straight from KNIME. The authoritative first 7 minutes before any third-party course.
Dify AI Tutorial: How To Use Dify AI to Build Apps (Beginner Guide 2026)
Best taxonomic intro: lets you map a problem to the right Dify pattern before you start building.
NEW Make.com AI Agents - 2026
Watch after the full guide when you want your Make scenarios to make judgment calls.
n8n Quick Start Tutorial: Build Your First AI Agent (2026)
The official source. If anything in another tutorial looks different from your actual n8n, this is the one whose UI matches today.
Make.com Tutorial for Beginners 2026 (Full Guide)
The one to bookmark. Metics Media's guide is the most-referenced Make tutorial; treat it as your lookup reference.
KNIME 08: KNIME Interface Explained: Nodes, Workflow & Execution | KNIME Tutorial | Data Analytics
The academic, no-rush walkthrough. Good if you want the interface mapped before cleaning real data.
Master Dify AI in 1 Hour: Complete Beginner's Guide for 2026
The best 'I have one evening' intro to Dify. By the end you'll know which of the four app patterns fits your use case.
How to Build Scalable Agentic RAG with Dify and Qdrant
Authoritative source on Dify's current positioning. Watch when you've outgrown 'just give me an answer' RAG and need agents that decide what to look up next.
+ 8 more in the video library.
9FAQ 34
What is Dify and what can I use it for?
Dify is an open-source platform for building AI-powered applications — like chatbots and question-answering assistants — without writing code. You design your app visually, connect an AI model (such as ChatGPT or Claude), and optionally upload documents for the AI to answer questions from. It is especially well-suited for non-technical users who want a working chatbot over their own documents in a short time.
Should I use Dify Cloud or self-host Dify — what is the difference?
Dify Cloud means signing up at dify.ai and using it in your browser with no installation — the easiest option for beginners. Self-hosting means running Dify on your own computer or server with Docker, which gives full data privacy and no usage limits beyond your hardware. Cloud is best for building quickly; self-hosting makes sense if your data is sensitive (e.g. unpublished research) or you need more documents and storage than the free cloud tier provides.
How does Dify compare to n8n — which should I use?
Dify is built specifically for AI chatbots and knowledge-base apps, with RAG, prompt engineering, and chat interfaces built in — you can have a working chatbot in under an hour with no coding. n8n is a general workflow-automation tool with hundreds of integrations that treats AI as one optional component; it has a steeper learning curve. For a chatbot over your papers, Dify is the right choice; n8n shines when you need to connect many external systems alongside AI.
What is a Dify Knowledge Base and how does RAG work?
A Knowledge Base is a collection of your own documents (papers, notes, PDFs) that you upload into Dify. When someone asks a question, Dify first searches those documents for the most relevant passages, then sends those passages along with the question to the AI model — a technique called Retrieval-Augmented Generation (RAG). The result is an AI that answers based on your specific materials rather than just its general training, making it far more accurate for specialized topics.
What file types can I upload to a Dify Knowledge Base?
Dify supports a wide range of formats: TXT, Markdown, PDF, HTML, Excel, Word, CSV, PowerPoint, and more, up to a per-file size limit. For most uploads of paper PDFs or Word notes, the standard mode works without extra setup; an alternative processing option unlocks the broader format list including presentations.
How do I upload my papers and notes to a Dify Knowledge Base step by step?
Go to the Knowledge section, click Create Knowledge, and name it. Drag-and-drop or browse to select your files, then choose automatic text chunking (recommended for beginners) and a high-quality indexing mode for better accuracy. Click Save and Process — indexing may take a minute or two. Once complete, attach this Knowledge Base to your chatbot app, and the bot can answer questions based on its contents.
What AI models does Dify support — can I use ChatGPT, Claude, Gemini, or local models?
All of them. Dify supports OpenAI (GPT), Anthropic (Claude), Google (Gemini), and many more cloud providers, and it also supports local models running on your own computer via Ollama — meaning you can run open-source models for free with no per-message cost, though that needs a capable machine. You add any provider under Settings → Model Providers by pasting in your API key.
What is KNIME and what is it used for?
KNIME Analytics Platform is a free, open-source desktop workbench for building data pipelines visually. Instead of writing code, you drag "nodes" onto a canvas and wire them together: one node reads a CSV, another filters rows, another trains a model, another draws a chart. Chained together, those nodes form a workflow that runs top-to-bottom and reproduces the same result every time. It is used for data cleaning, blending multiple sources, exploratory analysis, machine learning, and reporting. KNIME has a strong life-sciences pedigree and is widely used in drug discovery, NGS, and clinical data pipelines, but it suits any field that wrangles tabular data without a programming environment.
Does my data stay on my machine, or does it go to the cloud?
With the desktop KNIME Analytics Platform, your data stays local. Workflows execute on your own machine, files are read from and written to your own disk, and no account or upload is needed to build and run a pipeline. That makes it well suited to unpublished research or sensitive data that should not leave your environment. Data only goes to the cloud if you deliberately use a cloud connector (for example an S3 or Snowflake reader), publish a workflow to KNIME Community Hub, or call a hosted AI node such as OpenAI. The AI/LLM extension also lets you wire in local models via Ollama, keeping inference on-premises too.
What are KNIME's main limitations?
KNIME is memory-hungry: large datasets (roughly over 1 GB held in memory) can slow or crash the desktop app, and it is not a drop-in replacement for distributed engines like Spark or Dask out of the box. The visual canvas, which is a strength on small workflows, becomes hard to read once you have many branches and dozens of nodes ("node sprawl"); metanodes and components help by collapsing sections. Team sharing, scheduling, and web execution all require a paid Hub plan rather than the free desktop. And despite the no-code promise, discovering the right node among thousands and learning the configuration dialogs still carries a real initial learning curve.
How is KNIME different from a workflow tool like n8n?
Both let you wire nodes on a canvas, but they aim at different jobs. KNIME is a data-science workbench: its nodes read, clean, blend, model, and visualise tabular data, and a workflow runs as a batch analysis you execute and re-run for reproducible results. n8n is an automation and integration tool: its nodes connect apps and APIs and run on triggers or schedules to move data and fire actions between services. Reach for KNIME when the goal is analysing or modelling a dataset and producing a clean table, figure, or prediction; reach for n8n when the goal is automating a process that passes small payloads between SaaS tools. They overlap at data movement but optimise for opposite ends.
Is KNIME really free, and is it open source?
Yes. The desktop KNIME Analytics Platform is fully free and open source, and KNIME's own download page describes it as "a free and open source low-code/no-code software." You install it, build workflows, and run them locally with no licence fee and no usage limits on the desktop app. What costs money is the optional cloud collaboration layer (KNIME Community Hub paid plans and the enterprise Business Hub) used for team sharing, scheduling, and web deployment. For an individual learning data science or running analyses on their own machine, nothing about the core platform needs to be paid for.
How do I install KNIME and do I need an account?
Go to knime.com/downloads and pick the installer for your operating system. The download page shows a short registration form (email, company, location, role) that you complete before the download starts, but the installed Analytics Platform itself needs no sign-in: you build and run workflows locally with no account. A KNIME account is only required later if you want to publish workflows to KNIME Community Hub. After installing, launch the app, create a new workflow, and you are on an empty canvas ready to drag your first node from the node repository.
What operating systems does KNIME run on, and how much RAM do I need?
KNIME Analytics Platform runs on Windows (10, 11, and Server 2016/2019/2022), Linux (Ubuntu 20.04/22.04 LTS, RHEL/CentOS/Rocky 8 and 9), and macOS (Sonoma and Sequoia, on both Intel and Apple Silicon). It ships with the runtime it needs, so there is no separate Java install to manage. Memory is the setting most worth tuning: the default heap allocation is just 1024 MB, and KNIME recommends raising it to roughly half your available system RAM by editing the -Xmx value in the knime.ini file. On a 16 GB machine that means setting it to around 8 GB, which keeps larger tables responsive.
What does KNIME Hub cost, and what do the paid plans add?
KNIME Community Hub has a free Personal plan that connects to 300+ data sources, includes 20 K-AI interactions per month, and allows local sharing only. The Pro plan is $19/month (also EUR 19): it adds manual workflow runs, deployment as data apps, secrets storage, 500 K-AI interactions, and 120 execution credits (then $0.025 per vCore-minute) for one user. The Team plan is $99/month and adds private collaboration spaces for 3 members (extra members $49/month, up to 10 users) with centralised billing. Scheduling, REST-API deployment, and enterprise authentication live in Business Hub, which is priced on request. The desktop platform stays free regardless of which Hub tier you choose.
How do I build my first workflow in KNIME?
The smallest useful workflow is two nodes. Create a new workflow, drag a CSV Reader node from the node repository onto the canvas, and point it at a CSV file. Then drag a Row Filter node next to it and draw a connection from the CSV Reader's output port (the triangle on its right) to the Row Filter's input port. Double-click the Row Filter to configure which rows to keep, then press the green play button to execute. Right-click the Row Filter and open its output table to see your filtered rows. You have built a real, reproducible pipeline without writing a single line of code.
What do the red, yellow, and green lights on a node mean?
Every node carries a small traffic light at its base that tells you its state at a glance. Red means the node is not yet configured, so it cannot run. Yellow means it is configured and ready but has not executed. Green means it executed successfully and its output is available to the next node. The triangles on a node's sides are its ports: data enters on the left and leaves on the right. Reading these together is the core debugging skill in KNIME: a chain that stops at a yellow or red node shows you exactly where execution halted and which step still needs attention.
How do I combine or join two data files in KNIME?
Add a reader node for each source (for example a CSV Reader and an Excel Reader), point each at its file, and execute both so they turn green. To stack rows from two tables with the same columns, wire both into a Concatenate node. To match rows by a shared key (like an ID column), use a Joiner node instead, which performs inner, left, right, or outer joins. A Column Renamer node helps tidy mismatched column names so the tables line up first. Because each step is a visible node on the canvas, anyone opening the workflow can trace exactly how the messy inputs became one clean, analysis-ready table.
Can I do machine learning in KNIME without coding?
Yes. Machine learning is built from the same wired nodes as everything else. The standard pattern is: a Partitioning node splits your clean table into a training set and a test set (say 80/20); a Learner node (Decision Tree, Random Forest, or XGBoost Tree Ensemble) trains a model on the training set; and the matching Predictor node applies that model to the held-out test set. Wire a Scorer node onto the Predictor's output to get an accuracy figure and confusion matrix. To compare two algorithms fairly, put each on its own branch with its own Scorer and read the numbers side by side. The whole pipeline stays on the canvas, reproducible and shareable, with no code written.
What is Make and what can I use it for?
Make (formerly Integromat) is a visual automation platform that connects 3,000+ apps so they pass data between each other automatically, without code. You build a 'scenario' by dragging 'modules' (one per app or action) onto a canvas and wiring them together. A researcher could, for example, watch a folder for new PDFs, send each to an AI module to summarise, and append the result to a Google Sheet — running on a schedule with no manual steps.
+ 14 more in the library.
10Glossary 80 terms
Show the 80 terms
n8n
Workflow- A saved sequence of connected steps (nodes) that n8n runs automatically to move or transform data between apps.
Node- A single building block in a workflow — each node performs one action, such as sending an email, filtering data, or calling an API.
Trigger node- A special node that sits at the start of a workflow and decides when it runs — for example, on a schedule, when a form is submitted, or when another app sends a signal.
Action node- A node that does something in an external service — such as creating a row in Google Sheets, sending a Slack message, or reading an email.
Core node- A built-in utility node that handles data processing or flow control without connecting to an external service — examples include IF, Filter, Merge, and Code.
Connection- The arrow drawn between two nodes on the canvas that tells n8n to pass data from one node to the next when the workflow runs.
Canvas- The visual drag-and-drop workspace inside n8n where you build a workflow by placing and connecting nodes.
Execution- One complete run of a workflow — n8n records what happened at each node so you can inspect inputs, outputs, and any errors afterward.
Credentials- Securely stored login details (such as API keys or passwords) that let n8n connect to an external service on your behalf without exposing secrets inside the workflow.
Webhook- A URL that n8n creates for you so that an outside app can instantly start your workflow by sending data to that address.
Expression- A small piece of JavaScript written inside double curly braces ({{ }}) that lets you pull in data from a previous node or do a quick calculation instead of typing a fixed value.
Item- A single unit of data travelling through a workflow — for example, one email, one spreadsheet row, or one API result.
Schedule trigger- A trigger node that starts a workflow automatically at a set time or repeating interval, similar to a calendar alarm.
IF node- A core node that checks a condition and sends each data item down one of two paths — True or False — so different actions can happen depending on the data.
Sub-workflow- A separate workflow that another workflow calls like a reusable function, helping you keep complex automations organised and avoid repeating the same steps.
Error handling- A set of features in n8n — including dedicated error workflows and the Stop And Error node — that let you define what should happen if a node fails instead of silently stopping.
Data mapping- The act of telling a node where to find its input by dragging a field from a previous node's output onto the current node's input — no code required.
Sticky note- A text annotation you can place anywhere on the canvas to explain what part of a workflow does, without affecting how it runs.
Template- A pre-built workflow shared by the n8n community that you can import and adapt instead of building from scratch.
Manual execution- Running a workflow by clicking the Execute button yourself, used for testing before you switch the workflow on for automatic production runs.
Make
Scenario- One automation in Make: a trigger plus the connected modules that run when it fires.
Module- A single step in a scenario, representing one app action — read a row, send an email, call an API.
Bundle- One packet of data passing from one module to the next.
Router- A module that splits a scenario into several parallel routes.
Iterator- A module that loops over a list, running the later modules once per item.
Aggregator- A module that combines many bundles back into a single one.
Operation- One module action. Make meters usage by operations — now billed as "credits".
Credit- Make's unit of metered usage; each module action costs one (1,000/month on the free plan).
Make Code- A module for running custom JavaScript or Python when no ready-made app module fits.
Maia- Make's AI assistant — it builds and troubleshoots scenarios from a plain-language description.
Webhook- A URL that triggers a scenario the instant another service posts data to it.
Zapier
Zap- One automated workflow in Zapier: a trigger plus one or more actions.
Trigger- The event that starts a Zap — a new email, a new spreadsheet row, a form submission.
Action- A step a Zap performs after the trigger — create, update, or send something in another app.
Multi-step Zap- A Zap with more than one action chained together (a paid-plan feature).
Task- One action that completes successfully. Zapier bills by tasks, so each action that runs counts.
Path- Conditional branching — different actions run depending on the incoming data.
Filter- A step that stops a Zap from continuing unless its conditions are met.
Formatter- A built-in step that reshapes data — dates, text, numbers — between other steps.
Copilot- Zapier's AI that builds a Zap for you from a plain-English description.
Agents- Zapier AI agents that carry out multi-step tasks such as research or data lookups.
Tables- Zapier's built-in database for storing and reading data inside your automations.
KNIME
Workflow- A collection of connected nodes arranged on the editor canvas that together carry out a complete data analysis from reading in data to producing a result.
Node- A single building block in a KNIME workflow — a colored box that performs one task, such as reading a file, filtering rows, or training a model.
Node Repository- The searchable panel listing every node available in your KNIME installation; you drag nodes from here onto the workflow canvas.
Workflow Editor- The central canvas where you place and connect nodes to build a workflow.
Port- A connection point on the side of a node — input ports (left) receive data and output ports (right) send data to the next node.
Data Port- A port that passes a data table between nodes; shown as a black triangle and can only connect to another data port.
Model Port- A port that passes a trained machine-learning model from one node (e.g. a trainer) to another (e.g. a predictor).
Flow Variable- A named value (like a file path or a number) that travels between nodes through flow-variable ports and can change how a node is configured without editing it by hand.
Node Status- A traffic-light indicator beneath each node: red means not yet configured, yellow means configured and ready, green means successfully executed.
Component- A reusable, shareable group of nodes packaged as a single custom node with its own configuration dialog; can be published to KNIME Hub for others to use.
Metanode- A group of nodes collapsed into one box purely to keep the canvas tidy; unlike a Component it cannot be shared or given a custom dialog.
Workspace- The folder on your computer where KNIME stores all your workflows, node settings, and any data the workflow produces.
KNIME Hub- The online repository where you can store, share, and download KNIME workflows and components; available as a free Community Hub or an enterprise Business Hub.
K-AI- The built-in KNIME AI assistant that can answer questions about the platform (Q&A mode) or automatically extend your workflow by suggesting and adding nodes (Build mode).
Space Explorer- The file-browser panel in KNIME where you navigate and manage your workflows, folders, components, and data files.
Node Monitor- The bottom panel that shows the output table, statistics, or flow variables produced by a selected node after it has been executed.
Workflow Annotation- A free-text box you can place anywhere on the workflow canvas to document what a section of the workflow does; supports basic markdown formatting.
Columnar Backend- An optional KNIME execution engine that stores table data column-by-column using Apache Arrow, which reduces memory use on large datasets.
RowID- A unique identifier automatically assigned to every row in a KNIME data table, similar to a row number in a spreadsheet.
Execution- The act of running a node or an entire workflow so that it processes its input data and produces output; triggered by pressing F7 or clicking the Execute button.
Dify
Chatbot- The simplest Dify application type — you give it a model and a prompt, and users interact with it through a back-and-forth chat interface without needing tool calls or a multi-step workflow.
Agent- A Dify application type where the AI can reason, decompose tasks, and call tools on its own without you having to pre-define every step.
Workflow- A Dify application type that chains multiple steps (nodes) together in a visual diagram; it runs once from start to finish, taking input and returning a result through a fixed, repeatable process.
Chatflow- A Dify application type that combines a workflow's structured steps with a conversational chat interface, so each user message triggers the full node chain before a response is returned.
Node- A single building block inside a Dify workflow — each node does one job (call an LLM, run code, retrieve from a knowledge base, etc.) and passes its result to the next node.
LLM Node- A workflow node that sends a prompt to a language model — supporting text, images, and documents — and returns the model's output as a variable the rest of the workflow can use.
Knowledge Base- A repository of documents you upload to Dify so the AI can search and retrieve relevant passages when answering questions, rather than relying only on its training data.
RAG- Retrieval-Augmented Generation — a technique where the AI looks up relevant passages from your knowledge base and uses them to write a more accurate, grounded answer.
Chunking- The process of splitting an uploaded document into smaller pieces so each piece can be independently searched and retrieved by the AI.
Embedding- A way of converting text into numbers (vectors) so the system can measure how similar two pieces of text are — used internally when searching a knowledge base.
Reranking- A second sorting pass that takes the initial search results from a knowledge base and re-orders them by relevance using a scoring model so the best chunks reach the AI first.
ReAct- An agent strategy available in Dify that guides the model through explicit Thought → Action → Observation cycles, useful for models that lack native function-calling support.
System Prompt- Hidden instructions you write once that shape how the AI behaves in every conversation — setting its persona, tone, and rules before any user message arrives.
Variable- A named placeholder in a Dify prompt or workflow that gets filled in at run-time — either from user input or from the output of an earlier node.
Annotation- A hand-curated question-and-answer pair you save in Dify so the app returns your exact pre-written response whenever a sufficiently similar question appears, bypassing the LLM entirely.
Plugin- A modular component you install in your Dify workspace to extend it with new model providers, tools, or external integrations — usable across all apps in that workspace.
Tool- A specific action an Agent or workflow node can call — such as searching the web, running a calculation, or querying an API — to get information or take action beyond text generation.
API Key- A secret credential you enter in Dify when connecting a model provider or external service — it authenticates your workspace so the service accepts requests from your application.
11See also
💬 Discuss this chapter
Ask, share, or report — over on the Heidelberg AI community forum.