elias/claw-code
1
0
Fork 0
mirror of https://github.com/ultraworkers/claw-code.git synced 2026-04-26 22:47:38 +08:00
Code Issues Packages Projects Releases Wiki Activity

roadmap: #222 filed — Models list endpoint typed taxonomy is structurally absent: zero GET /v1/models and zero GET /v1/models/{id} surface across rust/crates/api/src/providers/anthropic.rs and rust/crates/api/src/providers/openai_compat.rs (rg returns zero hits for /v1/models, list_models, fetch_models, get_models, available_models, model_catalog, ModelInfo, ModelList, ListModelsResponse, OwnedBy, ModelObject, ModelCatalog across rust/), zero Model / ModelInfo / ModelList / ListModelsResponse typed taxonomy in rust/crates/api/src/types.rs, zero list_models<'a>(&'a self) -> ProviderFuture<'a, ModelList> and zero retrieve_model<'a>(&'a self, model_id: &'a str) -> ProviderFuture<'a, ModelInfo> methods on the Provider trait at rust/crates/api/src/providers/mod.rs:17-30 (only send_message and stream_message exist, both per-request), zero list_models dispatch on the ProviderClient enum at rust/crates/api/src/client.rs:8-14 (three variants Anthropic/Xai/OpenAi, all closed under per-request synchronous dispatch), zero claw models / claw model list / claw list-models CLI subcommand surface at rust/crates/rusty-claude-cli/src/main.rs, zero /models slash command in the SlashCommandSpec table at rust/crates/commands/src/lib.rs, zero validation against an authoritative source on set_model at rust/crates/rusty-claude-cli/src/main.rs:4989-5037 (user can type /model claude-banana-9000 and the runtime accepts it, swaps the active model to that string, and only fails at request time when the upstream provider returns 404 / invalid_model_error), and the existing /providers slash command at rust/crates/commands/src/lib.rs:716-720 is just a literal alias for /doctor at rust/crates/commands/src/lib.rs:1386-1389 despite advertising summary: "List available model providers" (advertised-but-rerouted shape — actively misleading at the UX layer, distinct from #220's advertised-but-unbuilt shape because the parse arm dispatches to a *different* command entirely instead of returning a clear unsupported error) — the canonical model-discovery affordance is invisible across every CLI / REPL / slash-command / Provider-trait / ProviderClient-enum / data-model surface, leaving claw-code's local hardcoded 13-entry MODEL_REGISTRY (3 anthropic + 5 grok + 1 kimi + 4 prefix routes for openai/gpt/qwen/kimi at rust/crates/api/src/providers/mod.rs:52-134 and 166-225) and its 6-entry model_token_limit match arm (rust/crates/api/src/providers/mod.rs:277-301 covering claude-opus-4-6, claude-sonnet-4-6, claude-haiku-4-5-20251213, grok-3, grok-3-mini, kimi-k2.5, kimi-k1.5 — returns None for current production IDs claude-opus-4-7, claude-haiku-4-6, gpt-5.2, o3, o4-mini, kimi-k3, qwen3-max, grok-4, deepseek-reasoner) as the only model-name knowledge the runtime has access to, with no way to refresh it, no way to discover new model IDs that providers publish, no way to validate user-supplied model strings, no way to cross-link to the pricing_for_model cost estimator (#209 substring-matching gap), no way to cross-link to the model_token_limit preflight check (#210 max_tokens shadow-fork gap silently no-ops on unknown models), no way to cross-link to the future is_batch_request flag (#221 batch-dispatch gap requires knowing which models support batch), and USAGE.md:426-440 documents only six model rows out of nine MODEL_REGISTRY entries (kimi alias missing from the documented table, four prefix routes mentioned only in passing prose, zero documentation of /v1/models endpoint usage / zero documentation of model-catalog discovery / zero documentation of "what to do when your provider ships a new model that isn't in claw-code's hardcoded registry") — the canonical model-discovery affordance is **the most universally-available endpoint in the LLM API ecosystem** (older than /v1/chat/completions itself, older than /v1/embeddings, older than /v1/messages, the literal first endpoint after auth on every OpenAI-compat provider since 2020 and on Anthropic since 2024-12-04, GA-shipped first-class typed surfaces in every Python/TypeScript SDK in the ecosystem) and claw-code is the **sole client/agent/CLI in the surveyed coding-agent ecosystem with zero /v1/models integration AND a misleading /providers slash command that aliases to /doctor** — both gaps are unique to claw-code in the surveyed ecosystem (Jobdori cycle #374 / extends #168c emission-routing audit / explicit follow-on candidate from #221's seven-layer-endpoint-family-absence shape — the third of three named candidates: Files API typed taxonomy / Embeddings API typed taxonomy / Models list endpoint typed taxonomy, and the most clawability-impacting because it's the upstream root cause of three downstream gaps already catalogued in this audit / sibling-shape cluster grows to twenty-one: #201/#202/#203/#206/#207/#208/#209/#210/#211/#212/#213/#214/#215/#216/#217/#218/#219/#220/#221/#222 / wire-format-parity cluster grows to twelve: #211+#212+#213+#214+#215+#216+#217+#218+#219+#220+#221+#222 / capability-parity cluster grows to four: #218+#220+#221+#222 / discovery-and-validation cluster: #222 alone but it's the upstream root cause of #209's pricing-fallback gap, #210's max_tokens shadow-fork gap, and #221's batch-dispatch gap / eight-layer-endpoint-family-absence-with-misleading-alias shape (endpoint-URL + data-model-taxonomy + Provider-trait-method + ProviderClient-enum-dispatch + CLI-subcommand-surface + slash-command-surface-with-misleading-alias + set_model-validation + downstream-consumers-with-stale-data) is the largest single advertised-vs-actual gap catalogued, distinct from prior single-field (#211/#212/#214) / response-only (#213/#207) / header-only (#215) / three-dimensional (#216) / classifier-leakage (#217) / four-layer (#218) / false-positive-opt-in (#219) / five-layer-feature-absence (#220) / seven-layer-endpoint-family-absence (#221) members; the advertised-but-rerouted shape is novel — strict-superset of #220's advertised-but-unbuilt because the parse arm dispatches to a *different* command instead of returning a clear unsupported error, applies to any future SlashCommandSpec entry where the summary field describes a feature different from what the parse arm dispatches to / external validation: Anthropic Models API reference at https://docs.anthropic.com/en/api/models-list documenting GET /v1/models GA 2024-12-04 with paginated before_id / after_id / limit and ModelInfo { id, type: "model", display_name, created_at } shape, Anthropic retrieve reference at https://docs.anthropic.com/en/api/models documenting GET /v1/models/{model_id} for single-model lookup, OpenAI Models API at https://platform.openai.com/docs/api-reference/models documenting the literal first endpoint after auth with Model { id, object: "model", created, owned_by } and ModelList { object: "list", data: Vec<Model> }, OpenAI Python SDK client.models.list() and client.models.retrieve(model_id) first-class typed surface, Anthropic Python SDK client.models.list() parallel surface GA-shipped 2024-12-04 alongside the API endpoint, Anthropic TypeScript SDK client.models.list(), AWS Bedrock ListFoundationModels API documenting Bedrock-anthropic-relay equivalent with FoundationModelSummary provider+model+modalities+active flag, Azure OpenAI Models reference with deployment-aware catalog, Vertex AI projects.locations.models.list for Vertex-published Anthropic/Gemini/3rd-party models, DeepSeek/Moonshot/Alibaba-DashScope/xAI parallel /v1/models OpenAI-compat shape, OpenRouter Models API at https://openrouter.ai/api/v1/models — the canonical "live model catalog with pricing" reference and the model that anomalyco/opencode-via-models.dev uses for pricing-data freshness, simonw/llm llm models and llm models default <model> first-class CLI subcommand backed by per-plugin model registration with models.dev-equivalent freshness, simonw/llm plugin-registration architecture for ad-hoc model addition, Vercel AI SDK 6 provider.languageModels() and provider.embeddingModels() first-class typed catalog APIs, LangChain init_chat_model(model_provider, model_name) reflective discovery via provider-defined catalogs and BaseChatModel.aget_models async catalog query, models.dev (https://models.dev) — community-maintained authoritative model catalog with pricing + capability flags + provider routing, used by anomalyco/opencode for pricing-data freshness with explicit fallback metadata when a model id isn't in the catalog (the canonical "external authoritative source for model metadata" reference), anomalyco/opencode models.dev integration with periodic refresh and explicit { provider: unknown, reason: not_in_pricing_table } fallback metadata, charmbracelet/crush typed catalog with provider+model+input/output-pricing, continue.dev config-file-driven catalog with auto-refresh from provider endpoints, zed-industries/zed bundled JSON catalog with periodic upstream refresh, TabbyML/tabby model catalog via plugin registration, llama.cpp server /v1/models local-model catalog via OpenAI-compat shape, LM Studio /v1/models local-model catalog, Ollama /api/tags and /v1/models local-model catalog with both Ollama-native and OpenAI-compat shapes, llamafile bundled-model catalog, LiteLLM models reference covering 100+ models at proxy level, portkey.ai gateway-level catalog, helicone.ai observability-platform model catalog with per-model usage stats, prompthub.us model-catalog-as-service, OpenTelemetry GenAI semconv gen_ai.request.model and gen_ai.response.model documented as required attributes for spans (every observability backend treats model as a first-class structured signal requiring authoritative-source validation), OpenAPI 3.1 spec for /v1/models at https://github.com/openai/openai-openapi as canonical machine-readable schema, Anthropic API stability versioning at https://docs.anthropic.com/en/api/versioning with anthropic-version header semver-stable since 2023-06-01 and models endpoint stable since 2024-12-04. Thirty-two ecosystem references, three first-class models-endpoint specs (Anthropic, OpenAI, OpenRouter), GA timeline of 16 months on Anthropic's side and 6+ years on OpenAI's side, eight first-class CLI/SDK implementations (Anthropic Python+TypeScript, OpenAI Python, simonw/llm, Vercel AI SDK, LangChain, Zed, charmbracelet/crush), seven first-class local-model catalogs (Ollama, LM Studio, llama.cpp server, llamafile, Tabby, Continue.dev, LiteLLM proxy), one community-maintained authoritative pricing source (models.dev) used by the closest peer coding agent. claw-code is the **sole client/agent/CLI in the surveyed coding-agent ecosystem with zero /v1/models integration AND a misleading /providers slash command that aliases to /doctor** — both gaps are unique to claw-code in the surveyed ecosystem, the model-discovery gap is the **upstream root cause** of three downstream cost-and-correctness gaps already catalogued in this audit (#209 / #210 / #221), and the misleading-alias-shape is novel within the cluster — #222 closes the upstream root cause of three downstream gaps and unblocks live-catalog-driven cost-estimation, max-tokens-validation, batch-capability-detection, and CLI-vs-slash-command-symmetry that the runtime's clawability doctrine treats as canonical baseline expectations.

Browse Source
This commit is contained in:
YeonGyu-Kim 2026-04-26 02:15:43 +09:00
parent 9acd4f14da
commit 0121f20a09
1 changed files with 294 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

294
ROADMAP.md
View File

File diff suppressed because one or more lines are too long