server-webui-and-protocols
Operate gptme server, Web UI, REST/SSE API, TUI, ACP, deployment, security, and protocol integration workflows.
- License 许可
- MIT
- 最近更新
- 2026-09-03
- Stars
- ★ 351
- Forks
- 27
01 前言 Front Matter
- name
- server-webui-and-protocols
- license
- MIT
- metadata
- { "disco-role": "operating" }
- description
- Operate gptme server, Web UI, REST/SSE API, TUI, ACP, deployment, security, and protocol integration workflows.
- disable-model-invocation
- true
02 资源文件 7
- 参考文档references/acp-and-tui.md7.2 KB
- 参考文档references/deployment.md2.9 KB
- 参考文档references/server-api.md15.9 KB
- 参考文档references/troubleshooting.md6.1 KB
- 参考文档references/webui-api.md9.4 KB
- 脚本scripts/check_webui_message_metadata.py13.3 KB
- 脚本scripts/probe_gptme_server.py8.2 KB
03 SKILL.md 正文
server-webui-and-protocols
Use this sub-skill for tasks about gptme-server, the REST/SSE API, the bundled or Vite Web UI, gptme-tui, ACP editor integration, deployment hardening, reverse proxies, or desktop sidecar behavior.
Keep this file as the router. Open the bundled references for details; do not depend on a target checkout's documentation when answering operational questions.
Route here for
- starting or probing
gptme-server, choosing server options, or checking server route availability - HTTP API, Server-Sent Events, auth cookies, bearer token handling, CORS, Host validation, or Local Network Access
- Web UI connection setup, multi-server settings, message metadata, streaming state, markdown rendering, or frontend/backend protocol mismatches
- ACP shim/direct/module launch, editor protocol lifecycle, stdout/stderr protocol failures, or server-side ACP step mode
gptme-tuiinstall, options, key limitations, or switching between TUI and CLI conversations- Docker Compose self-hosting, nginx reverse proxy, systemd service shape, public exposure hardening, or parent-death sidecar cleanup
Route away
- Generic terminal conversation usage, prompts, slash commands, log browsing, or
gptme-agentCLI: route tocli-and-conversations. - Provider keys, model selection, credential storage, provider plugins, or local model config: route to
configuration-and-providers. - Built-in tools, plugins, hooks, browser/computer tool internals, MCP, skills, or lessons: route to
tools-and-extensibility. - Evaluation suites, benchmark runs, SWE-bench/T-bench, Docker eval isolation, or leaderboards: route to
evals-and-benchmarks. - Maintainer-only checkout work such as frontend test selection, package release checks, lint/typecheck strategy, or code-review policy: route to
repo-development.
Read first
- references/server-api.md — install/serve options, safe help checks, auth/CORS/Host validation, route families, SSE events, message metadata, and
GptmeApiClient. - references/webui-api.md — bundled vs Vite Web UI, multi-backend behavior, frontend development gotchas, markdown paths, message metadata/SSE flow, Step grouping, and
ChatInputstate. - references/acp-and-tui.md — ACP shim/direct/module launch, protocol lifecycle, stdio failure modes, and TUI options/limitations.
- references/deployment.md — Docker Compose, nginx reverse proxy, systemd service pattern, self-hosting security, and sidecar parent-death behavior.
- references/troubleshooting.md — concrete symptoms, causes, and recovery steps for connection, streaming, metadata, ACP, TUI, and deployment failures.
Safe helper scripts
- scripts/probe_gptme_server.py probes a running server's root page,
/api/v2,/api/v2/version, and/api/v2/server/healthwith an optional bearer token. It performs only read-only requests and supports--help. - scripts/check_webui_message_metadata.py validates representative REST/SSE message dictionaries and compares the metadata fields expected by the Web UI. It supports JSON, JSON lines, and
data: {...}SSE samples and supports--help.
Fast operating checklist
- Identify the surface: local server, hosted Web UI, Vite frontend development, TUI, ACP, deployment, or sidecar cleanup.
- For browser connection failures, check token, exact CORS origin, and Local Network Access together; each is a separate gate.
- For message badge/rendering problems, compare REST conversation payloads with SSE
message_addedandgeneration_completepayloads before changing frontend code. - For ACP, treat stdout as JSON-RPC-only; logs and diagnostics must go to stderr.
- For persistent deployments, prefer a stable
GPTME_SERVER_TOKEN, loopback bind plus reverse proxy, disabled proxy buffering for SSE, and explicit parent-death watching for sidecars.
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).