configuration-and-providers
Configure gptme config files, credential storage, provider/model selection, local/custom providers, auth helpers, and optional dependency requirements.
- License 许可
- MIT
- 最近更新
- 2026-09-03
- Stars
- ★ 351
- Forks
- 27
01 前言 Front Matter
- name
- configuration-and-providers
- license
- MIT
- metadata
- { "disco-role": "operating" }
- description
- Configure gptme config files, credential storage, provider/model selection, local/custom providers, auth helpers, and optional dependency requirements.
- disable-model-invocation
- true
02 资源文件 5
- 参考文档references/configuration.md9.3 KB
- 参考文档references/providers-and-models.md13.5 KB
- 参考文档references/troubleshooting.md10.3 KB
- 脚本scripts/explain_model_selection.py15.5 KB
- 脚本scripts/validate_gptme_config.py26.0 KB
03 SKILL.md 正文
configuration-and-providers
Use this sub-skill when the task is about gptme configuration files, default model selection, provider prefixes, API-key storage, /account or gptme-auth, OpenRouter routing/privacy settings, local OpenAI-compatible backends such as Ollama, custom provider entries or provider packages, and optional extras/system dependency planning.
Route away when the task is mainly about:
- custom tools, plugin tools, hooks, MCP runtime behavior, skills, lessons, browser tools, or computer-use operation: use the tools/extensibility sub-skill.
gptme-serverauth tokens, CORS, Host validation, REST/SSE, Web UI, TUI, ACP, or deployment: use the server/protocols sub-skill.- eval model benchmarking, SWE-bench/T-bench, leaderboard generation, or pass-rate gates: use the evals/benchmarks sub-skill.
- maintainer-only changes inside a gptme checkout, test/lint/typecheck/doc/release workflows, or CI policy: use the repo-development sub-skill.
Read first
- references/configuration.md for global, local, project, and per-chat config files; merge/precedence rules; safe secret placement; model default conflicts; and static config review steps.
- references/providers-and-models.md for provider prefixes, API-key variables,
/account,gptme-auth, OpenRouter privacy/routing, local/Ollama setup, custom[[providers]], provider plugins, and extras/system dependencies. - references/troubleshooting.md for missing key errors, model-prefix mistakes, duplicate TOML keys, local model and summary-model issues, OpenRouter routing/privacy failures, OAuth/browser problems, and custom-provider pitfalls.
Safe helpers
- scripts/validate_gptme_config.py statically checks TOML config/env/provider setup without API calls and without printing secret values.
- scripts/explain_model_selection.py explains which model would win from supplied CLI, chat, config, and environment inputs without importing gptme or making network calls.
Example static passes:
python scripts/validate_gptme_config.py --help
python scripts/explain_model_selection.py --help
python scripts/explain_model_selection.py \
--models-default anthropic/claude-sonnet-4-6 \
--model-env openai/gpt-5 \
--api-key-env ANTHROPIC_API_KEY
Fast operating checklist
- Identify the scope: user-global config, project
gptme.toml, an existing chat log config, credentials/auth, or provider/model routing. - If both
[models].defaultandMODELare set, prefer[models].defaultfor the default chat model and treat the differingMODELvalue as a conflict to explain. - Keep secrets out of committable config. Put provider keys in
config.local.toml,gptme.local.toml, the credential store, or shell environment; never print the raw values. - Require provider-prefixed models for unambiguous selection:
anthropic/...,openai/...,openrouter/...,local/...,gptme/..., or a configured custom provider prefix. - For local Ollama/OpenAI-compatible backends, pair
local/<exact-model-tag>withOPENAI_BASE_URLand do not expectOPENAI_BASE_URLto affect non-local providers. - Prefer static checks and help commands (
gptme-auth --help,gptme-doctor --help) before launching browsers, making model calls, testing custom provider connectivity, or running checkout-maintainer tests.
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).