langchain-py
Build and maintain production-grade LangChain Python systems (LangChain 1.3.x / LangGraph 1.2.x baseline) with create_agent, middleware, tools, structured output, and multi-agent architectures (subagents, handoffs, router, skills). Activate for Python agent design, debugging, migrations from older A
v2026.11.13-21cd5683 installs+3 in 7 days39 StarsMITsynced 10 hours ago
Install
No NotHoly CLI yet? Install it first.
$ curl -fsSL https://notholy.com/install.sh | sh$ notholy skills install aeondave/langchain-pyWrites to ~/.notholy/skills after checking the signature and the SHA-256.
SignatureSigned, ed25519
Package4baca11f…16a35e52
4baca11fb0bf…9c1516a35e527 files · 6.8 KBScanLow1 finding
- Links to outside domains
Sourcegithub.com/AeonDave/malskill/tree/main/ai/langchain-py@21cd568
On this page
Practical workflow for building reliable LangChain systems with correct 1.x APIs and strong multi-agent design.
When to activate
- Building or refactoring Python agents with
create_agent - Designing multi-agent systems (subagents, handoffs, router, skills)
- Implementing middleware/context engineering/stateful tools
- Integrating Tavily or provider-specific tools
- Migrating snippets from stale pre-1.0 docs to current semantics
Version baseline (important)
Assume LangChain 1.3.x semantics (langchain-core on the 1.4.x line) with LangGraph 1.2.x unless the user explicitly requests another version.
- Prefer
from langchain.agents import create_agent - Prefer middleware-based dynamics (
dynamic_prompt,wrap_model_call,wrap_tool_call) - Use
Command(update=...)for state updates in tools/handoffs - Treat old
langgraph.prebuilt.create_react_agentsnippets as migration candidates — thelanggraph.prebuiltmodule is deprecated; canonical home islangchain.agents - Legacy chain abstractions (
LLMChain,initialize_agent,AgentExecutor, old memory classes) live inlangchain-classic;AgentExecutorreaches end of maintenance December 2026 — migrate tocreate_agentor a raw LangGraphStateGraph - LangGraph 1.2 adds per-node execution controls (timeouts, error-recovery policies, graceful shutdown hooks), the streaming API v3 (typed per-channel projections), and beta
DeltaChannelfor incremental checkpoint writes on large state
If uncertain, load references/version-scope.md first.
Recommended workflow
- Scope: identify version, providers, and required capabilities (single agent vs multi-agent).
- Choose architecture:
- Single agent for simple tasks and small toolsets
- Subagents for domain separation and centralized coordination
- Handoffs for sequential/user-facing state transitions
- LangGraph custom workflow for deterministic control gates
- Define tools with strong names, typed signatures, and clear docstrings.
- Engineer context via middleware (prompt/tool/model/response format selection).
- Add memory/persistence with checkpointer/store only where needed.
- Harden with error middleware, interrupts (HITL), and traceability.
- Verify version correctness (no legacy imports/patterns).
Multi-agent design rules
- Subagents: supervisor calls specialists as tools; great for domain boundaries.
- Handoffs: tool-driven state transitions; best for staged conversational workflows.
- Router: classification + dispatch for fast parallel domain fan-out.
- Skills: on-demand context loading while one main agent remains in control.
When implementing handoffs:
- Keep message history valid (tool-call and ToolMessage pairing).
- Use
Command.PARENTonly when explicit parent-graph routing is required. - Pass minimal context between agents; summarize rather than dumping internal traces.
Tavily integration guidance
Use Tavily as a focused retrieval/search tool, not as a blanket dependency for every agent.
- Add Tavily mainly to research subagents.
- Keep
max_resultsand depth bounded to control token/latency costs. - Use domain filters (
include_domains,exclude_domains) for precision.
Load references/tavily-integration.md for concrete patterns.
Anti-patterns
- Mixing pre-1.0 and 1.x APIs in same implementation
- Giving every agent every tool (routing degrades quickly)
- Handoffs without valid message/tool pairing
- Stateful workflows without checkpointer persistence
- Copy-pasting old blog snippets without migration audit
Outcome expectations
- Architecture choice is explicit and justified.
- Generated code is valid for LangChain 1.3.x / LangGraph 1.2.x.
- Multi-agent context boundaries are intentional and testable.
- Tooling and middleware are minimal but sufficient for reliability.
Resources
Load on demand:
references/version-scope.md— 1.3.x / LangGraph 1.2.x scope, migration-critical rules, changelog highlightsreferences/multi-agent-patterns.md— pattern selection and tradeoffsreferences/handoffs-and-command.md— state-machine handoffs,Command, message validityreferences/subagents-supervisor.md— supervisor/subagent layering and information flowreferences/tavily-integration.md— LangChain + Tavily setup and practical constraintsreferences/api-cheatsheet.md— fast API checklist forcreate_agent,ToolRuntime,Command
Files in this version
references/612.7 KB
- api-cheatsheet.md1.7 KB
- handoffs-and-command.md2.2 KB
- multi-agent-patterns.md2.3 KB
- subagents-supervisor.md2.0 KB
- tavily-integration.md1.4 KB
- version-scope.md3.1 KB
- SKILL.md5.0 KB
Permissions
| Permission | Declared | Found by the scan |
|---|---|---|
| Reads files | No | No |
| Writes files | Yes | Yes |
| Runs shell commands | Yes | Yes |
| Uses the network | Yes | Yes |
| Uses connectors | No | No |
| Drives a browser | No | No |
Scan findings
Reviewed on Nov 13, 2026
- InfoLink · Links to outside domains
A clean scan lowers the risk but does not prove a package is safe. Read what you install.
| Version | Published | Commit | SHA-256 | Signature | Permission change |
|---|---|---|---|---|---|
| v2026.11.13-21cd568 | Nov 13, 2026 | 21cd568 | 4baca11f | Signed | – |
More from AeonDave
All from this publisherMore in AI, agents and MCP
See the category- fal-assetsrehan-remade/fal-assets5
- ai-system-prompt-leakagenoorqureshi/ai-system-prompt-leakage4
- ai-prompt-injectionnoorqureshi/ai-prompt-injection4
- testing-prompt-injection-in-rag-pipelinesmukul975/testing-prompt-injection-in-rag-pipelines4
- implementing-llm-guardrails-for-securitymukul975/implementing-llm-guardrails-for-security4
- defending-llms-with-guardrailsmukul975/defending-llms-with-guardrails4
Ask anything your scope allows.
Create an account, add credits and ask the first question in the chat or the terminal.
- 1/3
Create the account
Individual or company. You can switch to a company later without losing balance or keys.
IndividualCompany - 2/3
Add credits
By card. Pay per token, no subscription.
$5.00minimum deposit - 3/3
Ask
In the chat, or install the CLI and run notholy login in your terminal.
$ notholy login