integrations-mcp-chainlit
Connect Langroid agents to MCP tools, web/search/file tools, Chainlit callbacks/UI, HTML logging, and external-service integrations.
- License 许可
- MIT
- 最近更新
- 2026-09-03
- Stars
- ★ 351
- Forks
- 27
01 前言 Front Matter
- name
- integrations-mcp-chainlit
- license
- MIT
- metadata
- { "disco-role": "operating" }
- description
- Connect Langroid agents to MCP tools, web/search/file tools, Chainlit callbacks/UI, HTML logging, and external-service integrations.
- disable-model-invocation
- true
02 资源文件 6
- 参考文档references/chainlit-and-ui.md5.9 KB
- 参考文档references/logging-and-output.md3.9 KB
- 参考文档references/mcp-workflows.md8.5 KB
- 参考文档references/search-and-file-tools.md5.9 KB
- 参考文档references/troubleshooting.md7.6 KB
- 脚本scripts/mcp_tool_smoke.py4.4 KB
03 SKILL.md 正文
integrations-mcp-chainlit
Use this sub-skill when a Langroid task needs to connect agents to external tools or service-facing integration layers: MCP servers, built-in web/search tools, safe local file tools, Chainlit UI callbacks, HTML task logs, quiet/status output control, or deterministic local smoke checks for those integrations.
Route before acting
- For Langroid custom
ToolMessagedesign, handler signatures, task done sequences, or multi-agent delegation mechanics, route to../agents-tasks-tools/SKILL.md. - For LLM provider configuration, model names, API gateways, or provider authentication, route to
../llm-provider-config/SKILL.md. - For RAG ingestion, document parsing, vector stores, or document-chat agents, route to
../retrieval-doc-chat/SKILL.md. - Stay here for external tool integration wiring, callback/UI integration, search/file tool enablement, MCP transport lifecycle, logging/output controls, and integration troubleshooting.
Operating map
- Identify which integration boundary is needed: MCP, search, file tools, Chainlit UI, or logging/output.
- Load the narrow reference for that boundary:
references/mcp-workflows.mdfor MCP tool generation, transports, async/decorator choices, and resource forwarding.references/search-and-file-tools.mdfor DuckDuckGo, Tavily, Exa, Google, Seltz, Twitter/X-over-MCP, and local file tools.references/chainlit-and-ui.mdfor Chainlit callback injection and UI session patterns.references/logging-and-output.mdfor HTML logs, quiet mode, streaming, and status output.references/troubleshooting.mdfor common event-loop, credential, subprocess, Chainlit, logging, and quiet-mode failures.
- Prefer async APIs inside any async app, Chainlit callback, notebook, or test: use
get_tool_async()/get_tools_async()and never wrap them inasyncio.run()from a running loop. - Keep external-service checks bounded: verify imports, environment variables, local
--helpavailability, and schemas before making live network calls. Do not require network for smoke tests. - When validating MCP availability without external side effects, run the bundled smoke script:
python scripts/mcp_tool_smoke.py
Minimal decision table
| Need | Preferred Langroid entry point | Reference |
|---|---|---|
| Build one MCP tool inside async code | await get_tool_async(server, "tool_name") |
references/mcp-workflows.md |
| Build all MCP tools | await get_tools_async(server) |
references/mcp-workflows.md |
| Customize MCP result formatting | subclass generated tool or use @mcp_tool with handle_async() |
references/mcp-workflows.md |
| Web search tool | enable DuckduckgoSearchTool, TavilySearchTool, ExaSearchTool, GoogleSearchTool, or SeltzSearchTool |
references/search-and-file-tools.md |
| Safe local file read/write/list | ReadFileTool.create, WriteFileTool.create, ListDirTool.create |
references/search-and-file-tools.md |
| Chainlit UI around an agent/task | ChainlitAgentCallbacks or ChainlitTaskCallbacks |
references/chainlit-and-ui.md |
| Suppress terminal noise or inspect logs | quiet_mode, async_stream_quiet, TaskConfig(enable_html_logging=...) |
references/logging-and-output.md |
04 同仓库更多 Skills
passnet-triton-opt
Make a single PassNet Triton kernel fast AND numerically faithful: performance model (when a replacement can win at all), block/grid/warp tuning, autotune policy, launch overhead, and per-op numeric recipes to pass the dtype baseline tolerances. Use when a pass matches and is correct but speedup below expected, or when correctness fails by small numeric margins.
passnet-solve
END-TO-END playbook for solving one PassNet sample: analyze the computation graphs, decide the optimization strategy (what to fuse, what to replace, what to leave alone), drive the iteration loop, and maximize the sample score. This is the ENTRY skill — invoke it first for any PassNet optimization task; it tells you when to use passnet-pattern-fusion, passnet-triton-opt and passnet-feedback.
passnet-skill
PassNet GPU kernel optimization via compiler passes. Design and implement Triton-based optimization passes, create pass files under ./pass_dir/, self-evaluate with pass_evaluator, and iterate to maximize GPU speedup.
passnet-pattern-fusion
Author PassNet pass files: write patterns that actually MATCH the FX graph, pick fusion regions (which ops to absorb into one Triton kernel), and structure multi-pass submissions with the mandatory shared-dispatch architecture. Use when creating or fixing pattern()/replacement_args()/replacement_func() files, when a pass "failed to match", or when deciding how to fuse multiple kernels/ops.
passnet-orchestrate
Round/batch planning and decision support for PassNet graph optimization. Use for multi-sample triage, worker allocation, eval-budget policy, and post-eval keep/revert decisions. For solving one concrete sample end to end, use passnet-solve as the entry skill; consult this skill only when you need broader planning or a specific decision gate.
passnet-feedback
Fast, deterministic feedback for PassNet work: pre-flight pattern/match verification WITHOUT burning a GPU evaluation, per-node bottleneck analysis of a sample's graphs, and eval-log parsing into per-variant status + estimated score + failure classification. Use BEFORE every GPU evaluation (check_pattern), at the START of a sample (analyze_graph), and AFTER every evaluation (parse_eval_log).