Preserve AWF model inventories with null endpoints and live reasoning metadata (G-7)

.prettierignore

@@ -21,3 +21,6 @@ schemas/logs-jsonl.schema.json
 
 .beads/
 .github/aw/
+
+# Recorded byte-exact AWF reflection fixtures (SHA-256 pinned)
+actions/setup/js/test_data/awf_reflect/*.json

actions/setup/js/test_data/awf_reflect/README.md

@@ -0,0 +1,17 @@
+# Recorded AWF `/reflect` fixtures
+
+These files are unchanged byte-for-byte copies of `sandbox/firewall/awf-reflect.json`
+from public `github/gh-aw` workflow-run artifacts. They are excluded from
+prettier so that their SHA-256 digests stay verifiable. Tests derive labelled
+pending-fetch and malformed variants in memory; they never edit these files.
+
+| File | Source run | Artifact | SHA-256 | Shape |
+| --- | --- | --- | --- | --- |
+| `run-38995410124-awf-reflect.json` | [smoke-claude-on-copilot 38995410124](https://github.com/github/gh-aw/actions/runs/38995410124), commit `12ba3ffcbaf5760b6094f52d8f7d61938e4e951e`, 2026-10-17 | `agent` (12087091651) | `9d59511015584d359b4159737d1d44ed12ad9426a1690e62148ab53b7543648a` | Five endpoints; configured Copilot with 40 models and 40 `routing_models`; four unconfigured endpoints with null `models` and `routing_models` |
+| `run-29812420778-legacy-awf-reflect.json` | [Smoke Copilot 29812420778](https://github.com/github/gh-aw/actions/runs/29812420778), commit `fed7b8398300613d6f79474d75d3ffcdc6665d18`, 2026-07-21 | `agent` (8488185681) | `da2570969e0d62bd10195f029b5472f750404b515784d8c68e9b267e0975b410` | Older AWF reflection without `routing_models`; configured Copilot with 36 models and four unconfigured endpoints with null `models` |
+
+The legacy file is the oldest retained public reflection found. It is not the
+source of a `pkg/cli/testdata/model_routing_golden/` case: those source runs are
+either private or expired (for example, run 37886511291 returns HTTP 404). Every
+AWF `/reflect` version since its introduction supplies endpoint `configured`, so
+no recorded reflection without `configured` exists to pin.

actions/setup/js/test_data/awf_reflect/run-29812420778-legacy-awf-reflect.json

@@ -0,0 +1 @@
+{"endpoints":[{"provider":"openai","port":10000,"base_url":"http://api-proxy:10000","configured":false,"models":null,"models_url":"http://api-proxy:10000/v1/models"},{"provider":"anthropic","port":10001,"base_url":"http://api-proxy:10001","configured":false,"models":null,"models_url":"http://api-proxy:10001/v1/models"},{"provider":"copilot","port":10002,"base_url":"http://api-proxy:10002","configured":true,"models":["claude-haiku-4.5","claude-opus-4.5","claude-opus-4.6","claude-opus-4.7","claude-opus-4.8","claude-sonnet-4.5","claude-sonnet-4.6","claude-sonnet-5","gemini-2.5-pro","gemini-3-flash-preview","gpt-3.5-turbo","gpt-3.5-turbo-0613","gpt-4","gpt-4-0613","gpt-4-o-preview","gpt-4.1","gpt-4.1-2025-04-14","gpt-4o","gpt-4o-2024-05-13","gpt-4o-2024-08-06","gpt-4o-2024-11-20","gpt-4o-mini","gpt-4o-mini-2024-07-18","gpt-5-mini","gpt-5.3-codex","gpt-5.4","gpt-5.4-mini","gpt-5.5","gpt-5.6-luna","gpt-5.6-sol","gpt-5.6-terra","mai-code-1-flash-picker","text-embedding-3-small","text-embedding-3-small-inference","text-embedding-ada-002","trajectory-compaction"],"models_url":"http://api-proxy:10002/models"},{"provider":"gemini","port":10003,"base_url":"http://api-proxy:10003","configured":false,"models":null,"models_url":"http://api-proxy:10003/v1beta/models"},{"provider":"vertex","port":10004,"base_url":"http://api-proxy:10004","configured":false,"models":null,"models_url":null}],"models_fetch_complete":true,"model_aliases":{"agent":["sonnet-6x","gpt-5.4","gpt-5.3","gemini-pro","any"],"any":["copilot/*","anthropic/*","openai/*","google/*","gemini/*"],"claude":["agent"],"codex":["agent"],"coding":["copilot/gpt-5*codex*","openai/gpt-5*codex*","gpt-5-codex","kimi"],"copilot":["agent"],"gemini":["agent"],"gemini-3-flash":["copilot/gemini-3*flash*","google/gemini-3*flash*","gemini/gemini-3*flash*"],"gemini-flash":["copilot/gemini-*flash*","google/gemini-*flash*","gemini/gemini-*flash*"],"gemini-pro":["copilot/gemini-*pro*","google/gemini-*pro*","gemini/gemini-*pro*"],"gpt-5":["copilot/gpt-5*","openai/gpt-5*"],"gpt-5-codex":["copilot/gpt-5*codex*","openai/gpt-5*codex*"],"gpt-5-mini":["copilot/gpt-5*mini*","openai/gpt-5*mini*"],"gpt-5.3":["copilot/gpt-5.3*","openai/gpt-5.3*"],"gpt-5.4":["copilot/gpt-5.4*","openai/gpt-5.4*"],"gpt-5.5":["copilot/gpt-5.5*","openai/gpt-5.5*"],"gpt-5.6":["copilot/gpt-5.6*","openai/gpt-5.6*"],"haiku":["copilot/*haiku*","anthropic/*haiku*"],"large":["fable","sonnet","gpt-5-pro","gpt-5","gemini-pro"],"mai-code":["copilot/MAI-Code*","copilot/mai-code*","openai/MAI-Code*"],"mai-code-1-flash-picker":["copilot/MAI-Code-1-Flash-picker*","copilot/mai-code-1-flash-picker*","openai/MAI-Code-1-Flash-picker*"],"mini":["haiku","gpt-5-mini","gpt-5-nano","gemini-flash-lite"],"opus":["copilot/*opus*","anthropic/*opus*"],"small":["mini"],"small-agent":["haiku","gpt-5-mini","gemini-flash"],"sonnet":["copilot/*sonnet*","anthropic/*sonnet*"],"sonnet-6x":["copilot/*sonnet-4.5*","copilot/*sonnet-4.6*","copilot/*sonnet-5*","copilot/*sonnet-4-5-*","anthropic/*sonnet-4-5-*","copilot/*sonnet-4-6*","anthropic/*sonnet-4-6*","anthropic/*sonnet-5*"],"summarization":["haiku","gpt-5-mini","gemini-flash-lite","mini"],"vision":["copilot/gemini-*image*","google/gemini-*image*","gemini/gemini-*image*","copilot/gemini-*flash*","google/gemini-*flash*","gemini/gemini-*flash*"]},"model_fallback":{"enabled":true,"strategy":"middle_power","excludeEngines":[]},"model_fallback_effective":{"openai":{"enabled":true,"strategy":"middle_power","excludeEngines":[],"suppressed":false},"anthropic":{"enabled":true,"strategy":"middle_power","excludeEngines":[],"suppressed":false},"copilot":{"enabled":false,"strategy":"middle_power","excludeEngines":[],"suppressed":true,"suppression_reason":"copilot_standard_authoritative"},"gemini":{"enabled":true,"strategy":"middle_power","excludeEngines":[],"suppressed":false},"vertex":{"enabled":true,"strategy":"middle_power","excludeEngines":[],"suppressed":false}},"ai_credits":{"total":225.64295,"by_model":{"gpt-5.4-2026-03-05":{"input_credits":188.21525,"cached_input_credits":17.0112,"cache_write_credits":0,"output_credits":20.4165,"total":225.64295}}},"runs":{"enabled":true,"max_runs":500,"invocation_count":18,"remaining_runs":482},"cache_misses":{"enabled":true,"max_cache_misses":5,"consecutive_cache_misses":0,"remaining_cache_misses":5},"permission_denied":{"enabled":false,"max_permission_denied":null,"denied_count":0},"model_api_mapping":{"available":true,"last_updated":"2026-07-16T06:03:36Z","providers":["openai","anthropic"],"models":{"openai":{"endpoints":{"responses":{"path":"/v1/responses","description":"Newer unified API for agentic, tool-using, and multimodal workflows. Required for GPT-5+ models; some older models (o1, o3, gpt-4o, gpt-4.1) support it alongside chat/completions."},"chat_completions":{"path":"/v1/chat/completions","description":"Legacy conversational API. Being deprecated for newer models by end of 2026."}},"models":[{"family":"gpt-5.6","patterns":["gpt-5.6*"],"endpoints":["chat_completions","responses"],"notes":"Supports both endpoints. Includes gpt-5.6-sol, gpt-5.6-terra, gpt-5.6-luna per OpenAI SDK ChatModel."},{"family":"gpt-5.5","patterns":["gpt-5.5*"],"endpoints":["responses"],"notes":"Responses API only. Not present in OpenAI OpenAPI spec v2.3.0; reserved for future gpt-5.5 family models."},{"family":"gpt-5.4","patterns":["gpt-5.4*"],"endpoints":["chat_completions","responses"],"notes":"Supports both endpoints on OpenAI. Includes gpt-5.4-mini, gpt-5.4-nano. NOTE: gpt-5.4-mini is not accessible via /chat/completions on the Copilot API (responses-only via Copilot); the api-proxy handles this by falling back to the next alias candidate."},{"family":"gpt-5.3","patterns":["gpt-5.3*"],"endpoints":["chat_completions","responses"],"notes":"Supports both endpoints. Only gpt-5.3-chat-latest confirmed in OpenAPI spec; gpt-5.3-codex variants not present."},{"family":"gpt-5.2","patterns":["gpt-5.2*"],"endpoints":["chat_completions","responses"],"notes":"Supports both endpoints. Includes gpt-5.2-pro."},{"family":"gpt-5.1-codex-max","patterns":["gpt-5.1-codex-max"],"endpoints":["responses"],"notes":"Responses API only (ResponsesOnlyModel)."},{"family":"gpt-5.1","patterns":["gpt-5.1*"],"endpoints":["chat_completions","responses"],"notes":"Supports both endpoints for gpt-5.1, gpt-5.1-codex, gpt-5.1-mini."},{"family":"gpt-5-codex/pro","patterns":["gpt-5-codex","gpt-5-pro","gpt-5-pro-*"],"endpoints":["responses"],"notes":"Responses API only for gpt-5-codex and gpt-5-pro (ResponsesOnlyModel). Includes dated variants like gpt-5-pro-2025-10-06."},{"family":"gpt-5","patterns":["gpt-5","gpt-5-*"],"endpoints":["chat_completions","responses"],"notes":"Supports both endpoints for gpt-5, gpt-5-mini, gpt-5-nano."},{"family":"o4-mini-deep-research","patterns":["o4-mini-deep-research","o4-mini-deep-research-*"],"endpoints":["responses"],"notes":"Responses API only (ResponsesOnlyModel). Includes dated variants like o4-mini-deep-research-2025-06-26."},{"family":"o4","patterns":["o4*"],"endpoints":["chat_completions","responses"],"notes":"Reasoning model. o4-mini supports both endpoints (ModelIdsShared)."},{"family":"o3-pro/deep-research","patterns":["o3-pro","o3-pro-*","o3-deep-research","o3-deep-research-*"],"endpoints":["responses"],"notes":"Responses API only. o3-pro and o3-deep-research are in ResponsesModel but not ChatModel."},{"family":"o3","patterns":["o3*"],"endpoints":["chat_completions","responses"],"notes":"Reasoning model. o3, o3-mini support both endpoints. o3-pro/deep-research use responses only (covered by separate entry above)."},{"family":"o1-pro","patterns":["o1-pro","o1-pro-*"],"endpoints":["responses"],"notes":"Responses API only. o1-pro is in ResponsesModel but not ChatModel."},{"family":"o1","patterns":["o1*"],"endpoints":["chat_completions","responses"],"notes":"Reasoning model. o1, o1-mini, o1-preview support both endpoints. o1-pro uses responses only (covered by separate entry above)."},{"family":"computer-use-preview","patterns":["computer-use-preview*"],"endpoints":["responses"],"notes":"Responses API only. Specialized model for computer use tasks."},{"family":"codex-mini","patterns":["codex-mini*"],"endpoints":["chat_completions","responses"],"notes":"Supports both endpoints. codex-mini-latest is in ChatModel."},{"family":"chatgpt-4o-latest","patterns":["chatgpt-4o-latest"],"endpoints":["chat_completions","responses"],"notes":"Rolling alias for the latest ChatGPT-optimized GPT-4o model. Supports both endpoints (ModelIdsShared). Not matched by gpt-4o* pattern due to chatgpt- prefix."},{"family":"gpt-4o","patterns":["gpt-4o*"],"endpoints":["chat_completions","responses"],"notes":"Multimodal. Supports both endpoints. Includes audio and search preview variants."},{"family":"gpt-4.1","patterns":["gpt-4.1*"],"endpoints":["chat_completions","responses"],"notes":"Supports both endpoints. Includes gpt-4.1-mini, gpt-4.1-nano."},{"family":"gpt-4.5","patterns":["gpt-4.5*"],"endpoints":["chat_completions"],"notes":"Preview model. Chat completions only."},{"family":"gpt-4-turbo","patterns":["gpt-4-turbo*"],"endpoints":["chat_completions","responses"],"notes":"Legacy. In ModelIdsShared per OpenAI spec v2.3.0; supports both endpoints."},{"family":"gpt-4","patterns":["gpt-4","gpt-4-*"],"endpoints":["chat_completions","responses"],"notes":"Legacy. In ModelIdsShared per OpenAI spec v2.3.0; supports both endpoints."},{"family":"gpt-3.5-turbo","patterns":["gpt-3.5-turbo*"],"endpoints":["chat_completions","responses"],"notes":"Legacy. In ModelIdsShared per OpenAI spec v2.3.0; supports both endpoints. Deprecated."}]},"anthropic":{"endpoints":{"messages":{"path":"/v1/messages","description":"Unified Messages API for all Claude models. Supports tool use, streaming, extended thinking, and vision."}},"models":[{"family":"claude-fable-5","patterns":["claude-fable-5*"],"endpoints":["messages"],"notes":"Latest flagship. 1M token context."},{"family":"claude-mythos-5","patterns":["claude-mythos-5*","claude-mythos-preview*"],"endpoints":["messages"],"notes":"Invitation-only (Project Glasswing). Limited availability. 1M token context."},{"family":"claude-sonnet-5","patterns":["claude-sonnet-5*"],"endpoints":["messages"],"notes":"Best combination of speed and intelligence. 1M token context."},{"family":"claude-opus-4","patterns":["claude-opus-4*"],"endpoints":["messages"],"notes":"Includes claude-opus-4-8, claude-opus-4-7, claude-opus-4-6, claude-opus-4-5, claude-opus-4-1. 1M token context."},{"family":"claude-sonnet-4","patterns":["claude-sonnet-4*"],"endpoints":["messages"],"notes":"Includes claude-sonnet-4-6, claude-sonnet-4-5. Production workhorse. 1M token context."},{"family":"claude-haiku-4","patterns":["claude-haiku-4*"],"endpoints":["messages"],"notes":"Includes claude-haiku-4-5. Fast/cheap. 200K token context."},{"family":"claude-3.5-sonnet","patterns":["claude-3-5-sonnet*","claude-3.5-sonnet*"],"endpoints":["messages"],"notes":"Previous generation. 200K token context."},{"family":"claude-3.5-haiku","patterns":["claude-3-5-haiku*","claude-3.5-haiku*"],"endpoints":["messages"],"notes":"Previous generation fast model."},{"family":"claude-3.7-sonnet","patterns":["claude-3-7-sonnet*","claude-3.7-sonnet*"],"endpoints":["messages"],"notes":"Previous generation extended thinking model. 200K token context."},{"family":"claude-3-opus","patterns":["claude-3-opus*"],"endpoints":["messages"],"notes":"Previous generation. 200K token context."}]}},"error":null}}

actions/setup/js/types/unified_session.d.ts

@@ -228,10 +228,23 @@ export interface ModelRoutingOutcomeData {
 export interface ModelInventoryData {
   endpoints: {
     provider: string;
-    models: string[];
+    /** AWF's endpoint configuration state; absent when the source did not supply it. */
+    configured?: boolean;
+    /** Advertised model IDs in source order; absent means unknown, not empty. */
+    models?: string[];
+    /** Routing metadata in source order; absent means unknown, not empty. */
     routingModels?: {
       modelId: string;
       supportedEndpoints?: string[];
+      /** Identifier-shaped reasoning efforts; absent means unknown; a supplied empty array stays empty. */
+      supportedReasoningEfforts?: string[];
+      /**
+       * Positive context window size in tokens; absent means unknown.
+       * @asType integer
+       * @minimum 1
+       * @maximum 9007199254740991
+       */
+      contextWindowTokens?: number;
     }[];
   }[];
   modelsFetchComplete?: boolean;

actions/setup/js/unified_session.test.cjs

@@ -186,6 +186,8 @@ describe("Unified conclusion session", () => {
     '{"providers":{"copilot":{"models":["a",false]}}}',
     '{"provider":"copilot","models":["a",42]}',
     '{"endpoints":[{"provider":"copilot","models":[],"routing_models":[{"model_id":42}]}]}',
+    '{"endpoints":[{"provider":"copilot","configured":true,"models":["a"],"routing_models":[{"model_id":"a","supported_reasoning_efforts":["very high"]}]}]}',
+    '{"endpoints":[{"provider":"copilot","configured":true,"models":["a"],"routing_models":[{"model_id":"a","context_window_tokens":0}]}]}',
   ])("warns once and skips malformed model inventories: %s", content => {
     write("sandbox/firewall/models.json", content);
     const warn = vi.fn();
@@ -232,7 +234,7 @@ describe("Unified conclusion session", () => {
     ],
     [{ providers: {} }, { endpoints: [] }],
     [{ provider: "copilot", models: [] }, { endpoints: [{ provider: "copilot", models: [] }] }],
-    [{ endpoints: [{ provider: "copilot", configured: false, models: null }] }, { endpoints: [{ provider: "copilot", models: [] }] }],
+    [{ endpoints: [{ provider: "copilot", configured: false, models: null }] }, { endpoints: [{ provider: "copilot", configured: false }] }],
   ])("keeps genuine empty or unconfigured preferred inventories instead of stale mirrors: %j", (inventory, expected) => {
     write("agent/awf-reflect.json", inventory);
     write("sandbox/firewall/awf-reflect.json", { endpoints: [{ provider: "stale", models: ["stale"] }] });