roadmodel 0.2.65__tar.gz → 0.2.68__tar.gz

{roadmodel-0.2.65 → roadmodel-0.2.68}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.5
 Name: roadmodel
-Version: 0.2.65
+Version: 0.2.68
 Summary: BYO-key CLI that recommends the right AI model, platform, and settings for a prompt.
 Project-URL: Homepage, https://roadmodel.ai
 Project-URL: Repository, https://github.com/nathanramoscfa/roadmodel

{roadmodel-0.2.65 → roadmodel-0.2.68}/docs/benchmarks.json

@@ -6,7 +6,7 @@
     "endpoint": "https://artificialanalysis.ai/api/v2/data/llms/models",
     "attribution": "Benchmark data © Artificial Analysis, used under its API terms with attribution."
   },
-  "generated_at_utc": "2026-10-02T22:10:28Z",
+  "generated_at_utc": "2026-10-06T22:11:12Z",
   "model_count": 54,
   "unmapped": [],
   "missing_slugs": [],
@@ -38,8 +38,8 @@
         "tau2": 0.546783625730994,
         "tau_banking": 0.0927835051546392
       },
-      "median_output_tokens_per_second": 116.991,
-      "median_time_to_first_token_seconds": 13.947
+      "median_output_tokens_per_second": 98.659,
+      "median_time_to_first_token_seconds": 7.086
     },
     "claude-fable-5": {
       "aa_id": "cd55210d-358e-4df1-ba9c-9acb5f186cc9",
@@ -96,8 +96,8 @@
         "tau2": null,
         "tau_banking": 0.472164948453608
       },
-      "median_output_tokens_per_second": 70.155,
-      "median_time_to_first_token_seconds": 137.867
+      "median_output_tokens_per_second": 69.251,
+      "median_time_to_first_token_seconds": 140.452
     },
     "claude-opus-5": {
       "aa_id": "b8fc61f7-5e9a-49e6-8547-6ac56db24627",
@@ -154,8 +154,8 @@
         "tau2": null,
         "tau_banking": null
       },
-      "median_output_tokens_per_second": 97.574,
-      "median_time_to_first_token_seconds": 449.114
+      "median_output_tokens_per_second": 96.823,
+      "median_time_to_first_token_seconds": 470.118
     },
     "claude-sonnet-5": {
       "aa_id": "23c86e4a-c769-43c0-a056-79e3cd15834f",
@@ -183,8 +183,8 @@
         "tau2": null,
         "tau_banking": 0.37319587628866
       },
-      "median_output_tokens_per_second": 92.634,
-      "median_time_to_first_token_seconds": 198.594
+      "median_output_tokens_per_second": 0,
+      "median_time_to_first_token_seconds": 0
     },
     "claude-sonnet-5-5": {
       "aa_id": "b171d979-5ec1-45de-b8b9-db1ee65e77ec",
@@ -212,8 +212,8 @@
         "tau2": null,
         "tau_banking": null
       },
-      "median_output_tokens_per_second": 140.474,
-      "median_time_to_first_token_seconds": 261.115
+      "median_output_tokens_per_second": 137.452,
+      "median_time_to_first_token_seconds": 269.765
     },
     "deepseek-flash": {
       "aa_id": "dbe7c625-3100-4463-b479-a228c41f75dd",
@@ -241,8 +241,8 @@
         "tau2": null,
         "tau_banking": null
       },
-      "median_output_tokens_per_second": 214.866,
-      "median_time_to_first_token_seconds": 0.836
+      "median_output_tokens_per_second": 211.532,
+      "median_time_to_first_token_seconds": 0.838
     },
     "deepseek-v4-pro": {
       "aa_id": "83173329-c09a-41f1-a028-a282a5f908d7",
@@ -270,8 +270,8 @@
         "tau2": null,
         "tau_banking": 0.395876288659794
       },
-      "median_output_tokens_per_second": 108.409,
-      "median_time_to_first_token_seconds": 1.137
+      "median_output_tokens_per_second": 94.405,
+      "median_time_to_first_token_seconds": 1.116
     },
     "gemini-2.5-flash": {
       "aa_id": "219ed587-60c5-4a48-9517-8480e08d0ca1",
@@ -386,8 +386,8 @@
         "tau2": 0.956140350877193,
         "tau_banking": 0.214432989690722
       },
-      "median_output_tokens_per_second": 123.648,
-      "median_time_to_first_token_seconds": 26.15
+      "median_output_tokens_per_second": 120.916,
+      "median_time_to_first_token_seconds": 16.499
     },
     "gemini-3.5-flash": {
       "aa_id": "0097ebf5-124f-42f6-9463-33b00e711f03",
@@ -502,8 +502,8 @@
         "tau2": null,
         "tau_banking": 0.449484536082474
       },
-      "median_output_tokens_per_second": 255.648,
-      "median_time_to_first_token_seconds": 11.765
+      "median_output_tokens_per_second": 193.352,
+      "median_time_to_first_token_seconds": 11.478
     },
     "glm-4.5-air": {
       "aa_id": "5d303dc9-c027-401f-9803-4e9aa3331007",
@@ -618,8 +618,8 @@
         "tau2": null,
         "tau_banking": 0.503092783505155
       },
-      "median_output_tokens_per_second": 73.964,
-      "median_time_to_first_token_seconds": 2.655
+      "median_output_tokens_per_second": 74.56,
+      "median_time_to_first_token_seconds": 2.696
     },
     "glm-5.3-flash": {
       "aa_id": "19496b81-9f41-4214-a77a-1df803b3c5ae",
@@ -647,8 +647,8 @@
         "tau2": null,
         "tau_banking": 0.472164948453608
       },
-      "median_output_tokens_per_second": 52.907,
-      "median_time_to_first_token_seconds": 3.043
+      "median_output_tokens_per_second": 48.776,
+      "median_time_to_first_token_seconds": 2.773
     },
     "gpt-5": {
       "aa_id": "48e50f00-1fd1-4acc-b337-61078aa341e6",
@@ -850,8 +850,8 @@
         "tau2": 0.859649122807018,
         "tau_banking": null
       },
-      "median_output_tokens_per_second": 98.35,
-      "median_time_to_first_token_seconds": 46.046
+      "median_output_tokens_per_second": 94.067,
+      "median_time_to_first_token_seconds": 34.541
     },
     "gpt-5.4": {
       "aa_id": "a89c4b28-2d8c-456e-88ea-255fb51fd2b6",
@@ -1053,8 +1053,8 @@
         "tau2": 0.862573099415205,
         "tau_banking": 0.402061855670103
       },
-      "median_output_tokens_per_second": 109.955,
-      "median_time_to_first_token_seconds": 94.607
+      "median_output_tokens_per_second": 103.757,
+      "median_time_to_first_token_seconds": 83.831
     },
     "gpt-6-astra": {
       "aa_id": "2f339a97-9a0d-499a-9cb5-e0db665bfa25",
@@ -1082,8 +1082,8 @@
         "tau2": null,
         "tau_banking": 0.414432989690722
       },
-      "median_output_tokens_per_second": 55.68,
-      "median_time_to_first_token_seconds": 197.034
+      "median_output_tokens_per_second": 50.735,
+      "median_time_to_first_token_seconds": 205.366
     },
     "gpt-6-luna": {
       "aa_id": "6fb13851-40ce-4bda-ad3f-6b38e0c5daa7",
@@ -1111,8 +1111,8 @@
         "tau2": null,
         "tau_banking": null
       },
-      "median_output_tokens_per_second": 138.11,
-      "median_time_to_first_token_seconds": 83.735
+      "median_output_tokens_per_second": 133.905,
+      "median_time_to_first_token_seconds": 68.659
     },
     "gpt-6-sol": {
       "aa_id": "c50ea08c-88c0-4eb7-85b8-27f2bbf6d527",
@@ -1140,8 +1140,8 @@
         "tau2": null,
         "tau_banking": null
       },
-      "median_output_tokens_per_second": 105.847,
-      "median_time_to_first_token_seconds": 105.231
+      "median_output_tokens_per_second": 0,
+      "median_time_to_first_token_seconds": 0
     },
     "gpt-6.1-sol": {
       "aa_id": "b25ac058-1e33-4c31-bbc9-a20c3ad0717a",
@@ -1169,8 +1169,8 @@
         "tau2": null,
         "tau_banking": null
       },
-      "median_output_tokens_per_second": 59.099,
-      "median_time_to_first_token_seconds": 170.489
+      "median_output_tokens_per_second": 57.423,
+      "median_time_to_first_token_seconds": 179.839
     },
     "gpt-oss-120b": {
       "aa_id": "f0083258-8646-45b8-8082-7aaf6c2ea82a",
@@ -1198,8 +1198,8 @@
         "tau2": 0.657894736842105,
         "tau_banking": 0.127835051546392
       },
-      "median_output_tokens_per_second": 156.511,
-      "median_time_to_first_token_seconds": 0.525
+      "median_output_tokens_per_second": 186.012,
+      "median_time_to_first_token_seconds": 0.502
     },
     "gpt-oss-20b": {
       "aa_id": "36f73aaf-d38a-4b56-a2b3-d04d17186910",
@@ -1227,8 +1227,8 @@
         "tau2": 0.60233918128655,
         "tau_banking": 0.0701030927835052
       },
-      "median_output_tokens_per_second": 200.593,
-      "median_time_to_first_token_seconds": 0.409
+      "median_output_tokens_per_second": 231.256,
+      "median_time_to_first_token_seconds": 0.412
     },
     "grok-4.3": {
       "aa_id": "948892b5-db03-4118-a4a8-ccd51ed871ea",
@@ -1314,8 +1314,8 @@
         "tau2": null,
         "tau_banking": null
       },
-      "median_output_tokens_per_second": 80.156,
-      "median_time_to_first_token_seconds": 19.497
+      "median_output_tokens_per_second": 69.431,
+      "median_time_to_first_token_seconds": 8.037
     },
     "kimi-k2.7-code": {
       "aa_id": "8d0cb231-7303-452c-9923-a9620b948475",
@@ -1343,8 +1343,8 @@
         "tau2": 0.900584795321637,
         "tau_banking": 0.202061855670103
       },
-      "median_output_tokens_per_second": 84.582,
-      "median_time_to_first_token_seconds": 1.233
+      "median_output_tokens_per_second": 83.243,
+      "median_time_to_first_token_seconds": 1.298
     },
     "kimi-k3": {
       "aa_id": "f7d2fc3e-1f7b-405f-818c-07952a4af78f",
@@ -1372,8 +1372,8 @@
         "tau2": null,
         "tau_banking": 0.45979381443299
       },
-      "median_output_tokens_per_second": 35.161,
-      "median_time_to_first_token_seconds": 2.906
+      "median_output_tokens_per_second": 45.073,
+      "median_time_to_first_token_seconds": 2.953
     },
     "mistral-large-3": {
       "aa_id": "4928e950-7f37-4475-b0dc-c5bad781a321",
@@ -1401,8 +1401,8 @@
         "tau2": 0.245614035087719,
         "tau_banking": 0.0577319587628866
       },
-      "median_output_tokens_per_second": 79.16,
-      "median_time_to_first_token_seconds": 0.723
+      "median_output_tokens_per_second": 79.072,
+      "median_time_to_first_token_seconds": 0.688
     },
     "mistral-medium-3.5": {
       "aa_id": "dd059b25-d82a-4ead-82a4-4adceaaec48b",
@@ -1430,8 +1430,8 @@
         "tau2": 0.941520467836257,
         "tau_banking": 0.150515463917526
       },
-      "median_output_tokens_per_second": 168.676,
-      "median_time_to_first_token_seconds": 0.564
+      "median_output_tokens_per_second": 162.625,
+      "median_time_to_first_token_seconds": 0.595
     },
     "mistral-small-4": {
       "aa_id": "3fd96175-4ef1-434c-8795-f873aec2abc1",
@@ -1459,8 +1459,8 @@
         "tau2": 0.412280701754386,
         "tau_banking": 0.0494845360824742
       },
-      "median_output_tokens_per_second": 173.831,
-      "median_time_to_first_token_seconds": 0.643
+      "median_output_tokens_per_second": 163.865,
+      "median_time_to_first_token_seconds": 0.542
     },
     "muse-spark-1.3": {
       "aa_id": "9999672c-b687-4026-9c15-5cef02ff53bb",
@@ -1488,8 +1488,8 @@
         "tau2": null,
         "tau_banking": 0.505154639175258
       },
-      "median_output_tokens_per_second": 163.786,
-      "median_time_to_first_token_seconds": 43.349
+      "median_output_tokens_per_second": 115.306,
+      "median_time_to_first_token_seconds": 26.027
     },
     "opus-4.7": {
       "aa_id": "e9a09db3-8fd6-41dd-ba2f-20e0a2bff7f2",

{roadmodel-0.2.65 → roadmodel-0.2.68}/docs/catalog.json

@@ -1,9 +1,9 @@
 {
   "schema_version": "4",
   "output_contract_version": 2,
-  "generated_at_utc": "2026-10-02T22:10:28Z",
+  "generated_at_utc": "2026-10-06T22:11:12Z",
   "source_doc_sha256": {
-    "model-selector.txt": "3a1897163e23e66afb77cf2f76ae84aa9354f85b07a235c27fb6bc55e7b657f8",
+    "model-selector.txt": "c18d9168548b6079ffe7e75cf891691b7ef3a2bfaa296dcd9a596f1e0a69e1ed",
     "model-tier-cost-scale.md": "2b3b0eb82237eb5cb5bf468a3fe971bcaf0160ad7d0bc700698d2c16640bc41d"
   },
   "models": [
@@ -24,7 +24,7 @@
         "speed": "S"
       },
       "jurisdiction": "us",
-      "headline_benchmarks": "AA Intelligence Index 16.9 (reasoning); Output Speed 122.5 tokens/s; AA-Omniscience -4.2; latency leader among Claude family",
+      "headline_benchmarks": "AA Intelligence Index 16.9 (reasoning); HLE 10.4%; SciCode 42.2; Terminal-Bench 4.0 0.0; AA-LCR 0.743; Output Speed 98.7 tokens/s; AA-Omniscience -4.2; latency leader among Claude family",
       "pricing_notes": "Hidden by default; Bedrock/Vertex: regional endpoints +10% surcharge; Cache: writes 1.25x, reads 0.1x",
       "best_for": "Speed-optimized lowest-cost Claude model, ideal for simple completions, high-volume repetitive tasks, and latency-sensitive workflows where a lightweight capable response matters more than deep reasoning",
       "superseded_by": null,
@@ -474,7 +474,7 @@
         "speed": "S"
       },
       "jurisdiction": "us",
-      "headline_benchmarks": "AA Intelligence Index 40.9 (high); HLE 47.8% (high); Terminal-Bench 2.1 87.6; Output Speed 237.7 tokens/s",
+      "headline_benchmarks": "AA Intelligence Index 40.9 (high); HLE 47.8%; SciCode 56.6; Terminal-Bench 4.0 19.7; AA-LCR 0.813; Terminal-Bench 2.1 87.6; Output Speed 193.4 tokens/s",
       "pricing_notes": "-",
       "best_for": "Frontier-class multimodal input, plus strong coding, planning, agentic work, long context and knowledge, with very fast output, at a low price. Its AA Intelligence Index of 40.9 ranks 18th of 54 measured models.",
       "superseded_by": null,

{roadmodel-0.2.65 → roadmodel-0.2.68}/docs/model-selector.txt

@@ -456,7 +456,19 @@
       default. It also makes Opus 4.7+ and Fable default to a 1M context
       window on Bedrock, Vertex, Foundry and the Claude apps gateway. The
       `/effort` vocabulary and the extended-thinking controls are
-      unchanged.
+      unchanged. 2.1.288 fixes resumed sessions dropping the model's
+      earlier thinking, retries thinking-only responses after a
+      mid-response API timeout, and makes a plugin-defined agent spawned
+      by name run with its own effort instead of the default. 2.1.289 is a
+      plugin / mod / permission-rule bug-fix release. Neither changes the
+      `/effort` vocabulary or the extended-thinking controls. 2.1.290
+      makes three effort/thinking fixes. A flagged message retried on a
+      fallback model with a different saved level no longer changes the
+      effort level. A reply that the output content filter stopped while
+      Claude was still thinking is now retried once. `/effort` sent from
+      `claude agents` to a busy background session now applies right
+      away. 2.1.291 is a hotfix for two regressions. Neither release
+      changes the `/effort` vocabulary or the extended-thinking controls.
     - OpenAI (Codex, OpenAI API, ChatGPT advanced controls):
       `reasoning_effort` knob — `low`, `medium`, `high`, `xhigh`, `max`,
       `ultra`. Codex's plan-mode `plan_mode_reasoning_effort` variant
@@ -1240,9 +1252,9 @@
              tier-coding="A" tier-planning="A" tier-agentic="A"
              tier-multimodal="S" tier-long-context="A" tier-knowledge="A"
              tier-speed="S"
-             headline-benchmarks="AA Intelligence Index 40.9 (high); HLE 47.8% (high); Terminal-Bench 2.1 87.6; Output Speed 237.7 tokens/s"
+             headline-benchmarks="AA Intelligence Index 40.9 (high); HLE 47.8%; SciCode 56.6; Terminal-Bench 4.0 19.7; AA-LCR 0.813; Terminal-Bench 2.1 87.6; Output Speed 193.4 tokens/s"
              pricing-notes="-"
-             prose-generated="best-for"
+             prose-generated="best-for headline-benchmarks"
              best-for="Frontier-class multimodal input, plus strong coding, planning, agentic work, long context and knowledge, with very fast output, at a low price. Its AA Intelligence Index of 40.9 ranks 18th of 54 measured models." />
       <model id="gpt-5.6-luna" name="GPT-5.6 Luna"
              input-price-per-1m="$0.20" output-price-per-1m="$1.20"
@@ -1269,8 +1281,9 @@
              tier-coding="C" tier-planning="C" tier-agentic="D"
              tier-multimodal="B" tier-long-context="A" tier-knowledge="D"
              tier-speed="S"
-             headline-benchmarks="AA Intelligence Index 16.9 (reasoning); Output Speed 122.5 tokens/s; AA-Omniscience -4.2; latency leader among Claude family"
+             headline-benchmarks="AA Intelligence Index 16.9 (reasoning); HLE 10.4%; SciCode 42.2; Terminal-Bench 4.0 0.0; AA-LCR 0.743; Output Speed 98.7 tokens/s; AA-Omniscience -4.2; latency leader among Claude family"
              pricing-notes="Hidden by default; Bedrock/Vertex: regional endpoints +10% surcharge; Cache: writes 1.25x, reads 0.1x"
+             prose-generated="headline-benchmarks"
              best-for="Speed-optimized lowest-cost Claude model, ideal for simple completions, high-volume repetitive tasks, and latency-sensitive workflows where a lightweight capable response matters more than deep reasoning" />
       <model id="muse-spark-1.3" name="Muse Spark 1.3"
              input-price-per-1m="$1.25" output-price-per-1m="$4.25"

{roadmodel-0.2.65 → roadmodel-0.2.68}/hatch_build.py

@@ -80,7 +80,15 @@ class BundleDocsHook(BuildHookInterface):  # type: ignore[misc]
         # source file: scripts/update_projects.py, which is also the bootstrap
         # URL, so it is mapped in rather than copied under src/.
         if self.target_name == "wheel":
-            updater = root / "scripts" / "update_projects.py"
-            if not updater.is_file():
-                raise FileNotFoundError(f"hatch_build.py: {updater} is missing")
-            force_include[str(updater)] = "roadmodel/update_projects.py"
+            for name in (
+                "update_projects",
+                "agent_context",
+                "agent_rules",
+                "upgrade_fleet",
+                "worktree_memory",
+                "updater_lock",
+            ):
+                updater = root / "scripts" / (name + ".py")
+                if not updater.is_file():
+                    raise FileNotFoundError(f"hatch_build.py: {updater} is missing")
+                force_include[str(updater)] = f"roadmodel/{name}.py"

{roadmodel-0.2.65 → roadmodel-0.2.68}/infra/README.md

@@ -153,7 +153,7 @@ is wired across three env scopes:
   Google Password Manager as `roadmodel ROADMODEL_IP_SALT`).
   Rotation cadence is **quarterly** per the Phase 7
   secrets-rotation policy. Defaults to a placeholder in
-  `web/lib/withRateLimit.ts` so local + CI builds work without
+  `web/lib/ip-salt.ts` so local + CI builds work without
   override, but production builds without the override silently
   bucket every IP under the same key, so the var **must** be
   set in every Vercel scope serving real traffic.
@@ -442,7 +442,9 @@ deliberate — Step 6 should not relitigate the Upstash decision.
 | `UPSTASH_REDIS_TOKEN`          | `roadmodel-web` Vercel env vars (Marketplace Upstash + alias) | Next.js rate limiter (live since Step 7 close-out)                          | web: production + preview + development + staging Custom Env (Step 7) | web: production + preview + development + staging |
 | `ROADMODEL_IP_SALT`            | `roadmodel-web` Vercel env vars (generated `openssl rand -hex 32`) | Next.js rate limiter daily IP+UA hashing salt (live since Step 7 close-out); rotate quarterly | web: production + preview + development + staging Custom Env (Step 7) | web: production + preview + development + staging |
 | `FRONTIER_ROADMAP_ENABLED`     | `roadmodel-web` Vercel env vars (default `false`)            | Next.js model-routing resolver — gates the Phase 5 Anthropic frontier branch | **not set anywhere in Phase 4 (defaults false)**         | web: production + preview + staging (set true on the scope where Phase 5 rolls out) |
-| `RECOMMEND_RATELIMIT_EXEMPT_USER_IDS` | `roadmodel-web` Vercel env vars (comma-separated founder Supabase uid; default empty) | Next.js rate limiter — user_ids that skip the `/api/recommend` daily limit (founder browser dogfooding); empty = nobody exempt | web: production (founder uid) | web: production (+ preview/staging if founder tests there) |
+| `RECOMMEND_RATELIMIT_EXEMPT_USER_IDS` | `roadmodel-web` Vercel env vars (comma-separated founder Supabase uid; default empty) | The FOUNDER list (`web/lib/funding-lane.ts`): these user_ids run `/api/recommend` and `/api/roadmap` on the operator's keys with no rate limit, and see the founder-only engines; empty = nobody | web: production (founder uid) | web: production (+ preview/staging if founder tests there) |
+| `RECOMMEND_INVITED_USER_IDS` | `roadmodel-web` Vercel env vars (comma-separated Supabase uids; default empty; personal data, never committed) | The INVITE list (`web/lib/funding-lane.ts`, Phase 4.11): these user_ids run on the operator's keys, 20 a day each; every caller on neither list gets `402 funding_required` before any paid call | web: production (Phase 4.11 Step 1) | web: production |
+| `ROADMODEL_DAILY_COST_CAP_USD` | `roadmodel-web` Vercel env vars (readable, not sensitive) | The operator lane's daily spend circuit breaker (`web/lib/spend-guard.ts`): once the UTC day's operator-funded `audit_log.cost_usd` reaches it, the operator lane answers 503 until midnight; 0 = off | web: production = `2` (Phase 4.11 Step 1) | web: production |
 
 Rules:
 

{roadmodel-0.2.65 → roadmodel-0.2.68}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "roadmodel"
-version = "0.2.65"
+version = "0.2.68"
 description = "BYO-key CLI that recommends the right AI model, platform, and settings for a prompt."
 readme = "README.md"
 requires-python = ">=3.11"
@@ -92,6 +92,11 @@ include = [
     "docs/catalog.json",
     "docs/benchmarks.json",
     "scripts/update_projects.py",
+    "scripts/agent_context.py",
+    "scripts/agent_rules.py",
+    "scripts/upgrade_fleet.py",
+    "scripts/worktree_memory.py",
+    "scripts/updater_lock.py",
     "hatch_build.py",
     "README.md",
     "LICENSE",