Dify x PUPU
Dify x PUPU
Dify x PUPU · Open, observable, extensible intelligent workflows and plugin ecosystem.
From Low‑Code to Pro‑Code: connect orchestration, engineering guardrails, ecosystem, triggers, and observability into a production loop.
| Part | Module | Focus |
|---|---|---|
| 01 | Breakthrough | Pain points · positioning · why now |
| 02 | Fusion | Visual DSL ↔ Pro‑Code · debugging view · collaboration |
| 03 | Extension | Plugins & Marketplace · tooling security · reuse |
| 04 | Connection | Triggers · types & matrix · hands-on cases |
| 05 | Observability | OpenTelemetry · multi-DB · cost/perf visibility |
| 06 | Summary | Evolution path · call to action |
Start from pain points; map Dify's role.
Top-100 GitHub repo by stars.
30 mins build: preprocess → retrieve → guardrail → return.
Chunk + clean, push vectors.
# python node\nchunks = split(doc, size=400)\nstore(chunks)
Top-k + safety filter.
Write back; Trace ID to APM/logs.
The surface is “missing pieces”; the root is “engineering gaps”.
One-stop LLM app platform: build, orchestrate, operate.
Workflow / Prompt Studio / RAG pipeline.
Retries, branching, quotas, env isolation.
Plugins & tools to connect external systems.
Tracing / token / cost visibility keeps prod stable.
The Tipping Point: Compute, Models, and Data align.
From deterministic programming to probabilistic engineering.
The "Operating System" for LLM apps, connecting models to business.
Visual DSL is an abstraction of pro engineering, not a toy.
Both camps need visualization, for different reasons:
| Low‑Code | Pro‑Code | |
|---|---|---|
| Role of visual | Programming language | Abstraction + debugging view |
| Primary goal | Lower cognitive load Explainable, teachable |
Understand complex DSLs faster Comparable, replayable |
| What it surfaces | Runnable logic | Engineering facts (Trace / Tokens / Variables) |
Same logic, two forms, co-evolving.
def handler(payload):
country = payload["country"]
score = risk_model.predict(payload)
if country == "US" and score < 0.2:
return {"route": "fast-lane"}
return {"route": "human-review", "score": score}Let visual DSL handle 80% glue logic; use pro-code for the 20% heavy lifting.
Branches, loops, RAG pipelines—align fast with biz needs.
Embed Python/Node scripts for data cleaning and math.
Trace IDs stitch logs; iterate prompts + code from evidence.
Node-level input/output diff with highlights.
Follow Trace ID to Jaeger / SLS.
trace_id=wf-1827 span=llm.call status=ok p95=180ms
Manage Prompts like Code: Versioning and Diffing.
Breaking silos between Prompt Engineers and Developers.
Beyond linear flows: Loops, Iterations, and Conditionals.
Marketplace & plugins make Dify platform-ready.
Standard APIs, auth, and multi-language runtimes.
Isolated execution with timeout and retry control.
API Key / OAuth / internal credentials.
Pluggable quotas and metering.
Marketplace publishing and versioning.
curl -X POST https://api.dify.ai/plugins/slack_bot \\
-H "Authorization: Bearer <token>" \\
-d '{
"channel_id": "ops-alert",
"text": "Build passed ✅",
"bot_token": "******"
}'171103.23plugin-9821 (jump to logs)manifest + schema
code node / runtime
local + sandbox
Marketplace / internal
"Definition is development": manifest + simple code.
bot_token, channel_id).Tool, send message via Slack SDK in _invoke.Trust no code. Isolation is key for platform stability.
3rd-party Plugin Code
Arbitrary Network Calls
High Resource Usage
Namespace Isolation
Resource Quotas
Network Whitelist
Agentic Behavior: Plugins are not just executors, but deciders.
def _invoke(self, params):
user_intent = analyze_intent(params['input'])
if user_intent == 'complex_task':
# Reverse call back to Dify API
response = dify_client.run_workflow(
workflow_id="wf-sub-task-001",
inputs={"query": params['input']},
user_id=params['user_id']
)
return {"result": response['data'], "delegated": True}
return {"result": simple_process(params), "delegated": False}
Connecting developers and enterprises for shared value.
Build high-quality tools
Publish to Marketplace
Revenue share
Distribution channel
Billing & Settlement
Security review
One-click install
SaaS Subscription
Solve business pain
Triggers move apps from passive Q&A to proactive execution.
Webhook → Dify parses payload → LLM routes → action out.
Receive external event, verify signature.
Choose model/path based on context.
Call plugins/tools, write output.
Tracing + token/cost.
Bring “when to start” back to the orchestration layer.
| Without triggers | With triggers |
|---|---|
| Scattered entrypoints: buttons / cron / direct calls | Unified entry: event source → mapping → routing |
| Hard to govern: no trace, hard to debug | Governable: replayable, observable, auditable |
| Start conditions live in scripts/code | Start conditions live in the workflow canvas |
| Trigger logic hard to reuse | One canvas: branch first, merge later |
Covers schedules, third-party events, and in-house systems.
Run at fixed cadence: reports, cleanup, health checks.
Third-party events: PR updates, ticket changes, docs edits. One subscription can drive multiple workflows.
For your systems. Unique URL per trigger; query/header/body map to variables; optional callback with workflow result.
Plugin trigger + LLM = automated code review loop.
action/pull_request/repo/sender.Ensuring zero event loss and non-blocking execution.
From email ingress to auto-classification and human loop.
Customer emails support@company.com → Webhook fires Dify.
Analyze content: "Refund", "Technical Issue", or "General"?
Draft reply → Send to Slack for approval → Click to send.
Transparent, traceable, portable engineering.
Dify ships fast—so probes must be non-invasive.
aliyun-instrument uses wrapt to hook into Dify's core methods at runtime.
WorkflowRunner.run) are swapped with a wrapper.wrapt ensures metadata (like __name__) is preserved, keeping it invisible to business logic.Per-span rates: LLM 20%, vector search 50%, critical transactions 100%.
Adjust prompt length + concurrency by queue depth and TPS.
Error >1% → sample 100% and enable debug logs.
Frontend looks fine but RAG returns empty; no app logs.
Workflow trace shows retriever runs fast and returns empty string.
Click "infrastructure link" on span to open SLS/Jaeger with TraceID.
One shard of vector DB timed out; retry policy missing. Add retry + fallback and redeploy.
DatabaseConfig adds PostgreSQL, MySQL, OceanBase.To ensure Schema compatibility across dialects, we introduced independent migration tests for MySQL.
Replacing dialect-specific Raw SQL with Python Helpers & ORM.
convert_datetime_to_date wraps truncation logic (Postgres
date_trunc vs MySQL date_format).
dataset_retrieval.py removes Raw SQL, using SQLAlchemy JSON
methods for type safety.
Granular operations: Know where every penny goes.
Unified orchestration + guardrails + ecosystem + triggers → observable, evolvable production AI.
