{"domain":"bay-run-mvp-zfmlsu2yla-uc.a.run.app","count":14,"changes":[{"captured_at":"2026-08-19T08:30:29","card_hash":"53d95ed614782795eb7d88480f7d5c0ce06e6de5c3a1a9669ab2b05598fd1429","previous_card_hash":"4540394696f063512c33fd63b9ec23d4a87a8f2bd8289c9ba799eb2ebcd391b6","diff":{"skills_added":[{"id":"free_verified_embeddings_interface_handoff","name":"Discover the free OpenAI-compatible embeddings API","description":"Return the landing page https://run.huggingbay.xyz/embeddings, OpenAI base URL https://run.huggingbay.xyz/v1, and exact embeddings, rerank, classification, status, and receipt-verification endpoints. No user task is executed through A2A.","tags":["free-embeddings-api","openai-compatible","reranking","classification","provenance-receipts"],"inputModes":["application/json","text/plain"],"outputModes":["application/json"]}],"skills_removed":[],"skills_changed":[],"fields_changed":[{"field":"description","before":"FREE during launch: every call below costs 0.00 USD. No card, no signup — mint a token and go. Public A2A activation and interface handoff for Bay Run. This A2A endpoint does not execute user tasks; REST or authenticated MCP performs one-call solve, bake-off, quote, run, and verify. Free during launch — no card, no signup; generous limits (60/min); pricing later with notice.","after":"FREE during launch: every call below costs 0.00 USD. No card, no signup — mint a token and go. Bay Run is the free, verified, OpenAI-compatible embeddings API, with warm CPU reranking and classification beside it. This public A2A endpoint does not execute user tasks; it hands clients to the exact REST, OpenAI-compatible, or authenticated MCP interfaces for embeddings, bake-off, quote, run, and verification. Free during launch — no card, no signup; generous limits (60/min); pricing later with notice."}],"other_changed":true,"is_empty":false,"human_summary":"added 1 skill · description FREE during launch: every call below cos → FREE during launch: every call below cos"}},{"captured_at":"2026-08-18T17:53:13","card_hash":"4540394696f063512c33fd63b9ec23d4a87a8f2bd8289c9ba799eb2ebcd391b6","previous_card_hash":"4753a75733cbc6853ff6e000d31119a9f6c0294c53b0f1633883f816fd5c2d09","diff":{"skills_added":[],"skills_removed":[],"skills_changed":[],"fields_changed":[],"other_changed":true,"is_empty":false,"human_summary":"internal fields changed"}},{"captured_at":"2026-08-18T09:21:39","card_hash":"4753a75733cbc6853ff6e000d31119a9f6c0294c53b0f1633883f816fd5c2d09","previous_card_hash":"57734d50aab8d0e67d7570b6f692359b751404f4f8099da715aa4f665745a9d2","diff":{"skills_added":[],"skills_removed":[],"skills_changed":[{"id":"small_model_bakeoff_interface_handoff","fields":["description"]}],"fields_changed":[{"field":"description","before":"FREE during launch: every call below costs 0.00 USD. No card, no signup — mint a token and go. Public A2A activation and interface handoff for Bay Run. This A2A endpoint does not execute user tasks; REST or authenticated MCP performs one-call bake-off, quote, run, and verify. Free during launch — no card, no signup; generous limits (60/min); pricing later with notice.","after":"FREE during launch: every call below costs 0.00 USD. No card, no signup — mint a token and go. Public A2A activation and interface handoff for Bay Run. This A2A endpoint does not execute user tasks; REST or authenticated MCP performs one-call solve, bake-off, quote, run, and verify. Free during launch — no card, no signup; generous limits (60/min); pricing later with notice."}],"other_changed":true,"is_empty":false,"human_summary":"updated 1 skill · description FREE during launch: every call below cos → FREE during launch: every call below cos"}},{"captured_at":"2026-08-18T01:26:11","card_hash":"57734d50aab8d0e67d7570b6f692359b751404f4f8099da715aa4f665745a9d2","previous_card_hash":"a98b0e5d3b4b89b125bbf2cfb53bc1c6bf104b43ceaa93da82b5a7f34c0bb4bc","diff":{"skills_added":[{"id":"small_model_bakeoff_interface_handoff","name":"Discover the Bay Run one-call bake-off","description":"Return the authenticated REST/MCP endpoints for one-call model bake-off, quote, run, and verification. No user task is executed through A2A.","tags":["task-routing","specialist-models","verification","mcp"],"inputModes":["application/json","text/plain"],"outputModes":["application/json"]}],"skills_removed":[{"id":"specialist_task_interface_handoff","name":"Discover the Bay Run task interface","description":"Return the authenticated REST/MCP endpoints for quote, run, and verification. No user task is executed through A2A.","tags":["task-routing","specialist-models","verification","mcp"],"inputModes":["application/json","text/plain"],"outputModes":["application/json"]}],"skills_changed":[],"fields_changed":[{"field":"description","before":"Public A2A activation and interface handoff for Bay Run. This A2A endpoint does not execute user tasks; authenticated REST or MCP performs quote, run, and verify.","after":"FREE during launch: every call below costs 0.00 USD. No card, no signup — mint a token and go. Public A2A activation and interface handoff for Bay Run. This A2A endpoint does not execute user tasks; REST or authenticated MCP performs one-call bake-off, quote, run, and verify. Free during launch — no card, no signup; generous limits (60/min); pricing later with notice."}],"other_changed":true,"is_empty":false,"human_summary":"added 1 skill · removed 1 skill · description Public A2A activation and interface hand → FREE during launch: every call below cos"}},{"captured_at":"2026-08-17T06:35:13","card_hash":"a98b0e5d3b4b89b125bbf2cfb53bc1c6bf104b43ceaa93da82b5a7f34c0bb4bc","previous_card_hash":"6999e07d373aead29aac1fb87be1855d789f1114dba236916061899f4b4c2ff3","diff":{"skills_added":[],"skills_removed":[],"skills_changed":[],"fields_changed":[],"other_changed":true,"is_empty":false,"human_summary":"internal fields changed"}},{"captured_at":"2026-08-16T02:06:24","card_hash":"6999e07d373aead29aac1fb87be1855d789f1114dba236916061899f4b4c2ff3","previous_card_hash":"87a4de1c0cc5fac8ceae2fa8864c00d26f1daba8af61420506f2b5ae1f563f57","diff":{"skills_added":[],"skills_removed":[],"skills_changed":[],"fields_changed":[],"other_changed":true,"is_empty":false,"human_summary":"internal fields changed"}},{"captured_at":"2026-08-15T07:40:01","card_hash":"87a4de1c0cc5fac8ceae2fa8864c00d26f1daba8af61420506f2b5ae1f563f57","previous_card_hash":"7f0cdd128fecd7ce1ec689bdea371ed6d0b7d0c0d33a032c78818741268a265e","diff":{"skills_added":[],"skills_removed":[],"skills_changed":[],"fields_changed":[],"other_changed":true,"is_empty":false,"human_summary":"internal fields changed"}},{"captured_at":"2026-08-14T09:36:16","card_hash":"7f0cdd128fecd7ce1ec689bdea371ed6d0b7d0c0d33a032c78818741268a265e","previous_card_hash":"6c52f896155040e12f946c87f7f1c3f911b7b83e94e0e79f0861637fc9a417ef","diff":{"skills_added":[{"id":"specialist_task_interface_handoff","name":"Discover the Bay Run task interface","description":"Return the authenticated REST/MCP endpoints for quote, run, and verification. No user task is executed through A2A.","tags":["task-routing","specialist-models","verification","mcp"],"inputModes":["application/json","text/plain"],"outputModes":["application/json"]}],"skills_removed":[{"id":"specialist_task_execution","name":"Quote, run, and verify a specialist task","description":"Discover the three-action REST/MCP handoff for constrained specialist execution.","tags":["task-routing","specialist-models","verification","mcp"],"inputModes":["application/json","text/plain"],"outputModes":["application/json"]}],"skills_changed":[],"fields_changed":[{"field":"description","before":"Decision and execution layer for specialized agent tasks: quote a qualifying route, execute under hard price and policy bounds, and verify the result.","after":"Public A2A activation and interface handoff for Bay Run. This A2A endpoint does not execute user tasks; authenticated REST or MCP performs quote, run, and verify."}],"other_changed":true,"is_empty":false,"human_summary":"added 1 skill · removed 1 skill · description Decision and execution layer for special → Public A2A activation and interface hand"}},{"captured_at":"2026-08-13T08:25:12","card_hash":"6c52f896155040e12f946c87f7f1c3f911b7b83e94e0e79f0861637fc9a417ef","previous_card_hash":"d54c7a96c0bf016595e99529c5d2bb0b7f52b61d321ab8bb6d5adc8a0de31b73","diff":{"skills_added":[{"id":"specialist_task_execution","name":"Quote, run, and verify a specialist task","description":"Discover the three-action REST/MCP handoff for constrained specialist execution.","tags":["task-routing","specialist-models","verification","mcp"],"inputModes":["application/json","text/plain"],"outputModes":["application/json"]}],"skills_removed":[{"id":"try_bay_run","name":"Try Bay Run","description":"Run the fixed no-argument Bay Run activation proof and receive connection next steps.","tags":["activation","specialist-models","mcp"],"inputModes":["application/json","text/plain"],"outputModes":["application/json"]}],"skills_changed":[],"fields_changed":[{"field":"description","before":"Run a fixed zero-credential activation proof, then connect to Bay Run's authenticated MCP or REST surfaces to parse PDFs/documents for RAG and discover, prove, and serve specialist models.","after":"Decision and execution layer for specialized agent tasks: quote a qualifying route, execute under hard price and policy bounds, and verify the result."},{"field":"version","before":"1.7.5","after":"2.0.0"}],"other_changed":true,"is_empty":false,"human_summary":"added 1 skill · removed 1 skill · description Run a fixed zero-credential activation p → Decision and execution layer for special · version 1.7.5 → 2.0.0"}},{"captured_at":"2026-08-12T07:20:25","card_hash":"d54c7a96c0bf016595e99529c5d2bb0b7f52b61d321ab8bb6d5adc8a0de31b73","previous_card_hash":"435e704ed4ed0fe6750df54c4c40c2e1e444b3eda7de3f4c1fae70570804f89e","diff":{"skills_added":[],"skills_removed":[],"skills_changed":[],"fields_changed":[{"field":"description","before":"Run a fixed zero-credential activation proof, then connect to Bay Run's authenticated MCP or REST surfaces to parse PDFs for RAG and discover, prove, and serve specialist models.","after":"Run a fixed zero-credential activation proof, then connect to Bay Run's authenticated MCP or REST surfaces to parse PDFs/documents for RAG and discover, prove, and serve specialist models."},{"field":"version","before":"1.4.0","after":"1.7.5"}],"other_changed":false,"is_empty":false,"human_summary":"description Run a fixed zero-credential activation p → Run a fixed zero-credential activation p · version 1.4.0 → 1.7.5"}},{"captured_at":"2026-08-11T11:30:45","card_hash":"435e704ed4ed0fe6750df54c4c40c2e1e444b3eda7de3f4c1fae70570804f89e","previous_card_hash":"d47f3685be56f8cf8760206217d09b325f177daf9e449a3c15d1c34a77f23268","diff":{"skills_added":[],"skills_removed":[],"skills_changed":[],"fields_changed":[{"field":"version","before":"1.2.0","after":"1.4.0"}],"other_changed":false,"is_empty":false,"human_summary":"version 1.2.0 → 1.4.0"}},{"captured_at":"2026-08-10T19:22:14","card_hash":"d47f3685be56f8cf8760206217d09b325f177daf9e449a3c15d1c34a77f23268","previous_card_hash":"0bdf80899d8b346d8052aae17e64dcdead38dadd3595ac75a9c4969f6378549e","diff":{"skills_added":[],"skills_removed":[],"skills_changed":[],"fields_changed":[{"field":"description","before":"Run a fixed zero-credential activation proof, then connect to Bay Run's authenticated MCP or OpenAI-compatible surfaces to discover, prove, and serve specialist models.","after":"Run a fixed zero-credential activation proof, then connect to Bay Run's authenticated MCP or REST surfaces to parse PDFs for RAG and discover, prove, and serve specialist models."},{"field":"version","before":"1.1.1","after":"1.2.0"},{"field":"url","before":"https://bay-run-mvp-889989800693.us-central1.run.app/a2a","after":"https://run.huggingbay.xyz/a2a"},{"field":"documentationUrl","before":"https://bay-run-mvp-889989800693.us-central1.run.app/llms-full.txt","after":"https://run.huggingbay.xyz/llms-full.txt"}],"other_changed":true,"is_empty":false,"human_summary":"description Run a fixed zero-credential activation p → Run a fixed zero-credential activation p · version 1.1.1 → 1.2.0 · url https://bay-run-mvp-889989800693.us-cent → https://run.huggingbay.xyz/a2a · documentationUrl https://bay-run-mvp-889989800693.us-cent → https://run.huggingbay.xyz/llms-full.txt"}},{"captured_at":"2026-08-10T11:09:36","card_hash":"0bdf80899d8b346d8052aae17e64dcdead38dadd3595ac75a9c4969f6378549e","previous_card_hash":"c51cca7d6833f2b642e4673899e5ccc8bf4775898c6875e41314c09097302712","diff":{"skills_added":[{"id":"try_bay_run","name":"Try Bay Run","description":"Run the fixed no-argument Bay Run activation proof and receive connection next steps.","tags":["activation","specialist-models","mcp"],"inputModes":["application/json","text/plain"],"outputModes":["application/json"]}],"skills_removed":[{"id":"calculate","name":"calculate","description":"Exact, instant arithmetic/math evaluation (no code exec — a hardened whitelist parser, not eval). + - * / // % **, parentheses, sqrt/log/sin/exp/floor/factorial, pi/e. The calculator agents need because LLMs botch multi-step arithmetic. Sub-millisecond.","tags":["specialist-models","inference","mcp"],"inputModes":["application/json"],"outputModes":["application/json"]},{"id":"classify","name":"classify","description":"Classify text with a small CPU-served specialist — the guardrail / safety / moderation / sentiment / intent / NLI layer agents need. FIXED-LABEL (any HF sequence-classification id, e.g. prompt-injection or toxicity) OR ZERO-SHOT (pass candidate_labels + an NLI model like facebook/bart-large-mnli, or model='auto'). Returns {labels:[{label,score}]}.","tags":["specialist-models","inference","mcp"],"inputModes":["application/json"],"outputModes":["application/json"]},{"id":"discover_models","name":"discover_models","description":"Search a 147K-model catalog for candidate small open specialist models for a narrow task (embedding|llm|vision|audio|tool|agent|any), ranked mirrored-first. Candidates, NOT proven winners — pass them to eval_models.","tags":["specialist-models","inference","mcp"],"inputModes":["application/json"],"outputModes":["application/json"]},{"id":"embed","name":"embed","description":"Turn text into embedding vectors with ANY open embedding / sentence-transformers id, served instantly on demand (OpenAI /v1/embeddings-compatible). Drops into any RAG / semantic-search pipeline; cheaper per-call than a frontier API at volume.","tags":["specialist-models","inference","mcp"],"inputModes":["application/json"],"outputModes":["application/json"]},{"id":"eval_models","name":"eval_models","description":"Prove which candidate wins on YOUR data — a head-to-head bake-off, not a public leaderboard (MTEB rank does not predict your-domain fit). Returns the single winner model id. task = embedding|rerank|extraction|generation.","tags":["specialist-models","inference","mcp"],"inputModes":["application/json"],"outputModes":["application/json"]},{"id":"extract","name":"extract","description":"Turn messy HTML/text (e.g. a scraper/Firecrawl dump) into schema-guided STRUCTURED JSON using a small CPU-served generative specialist. BEST-EFFORT: returns {data, json_valid, raw} — always CHECK `json_valid` before trusting `data`. Deterministic (greedy).","tags":["specialist-models","inference","mcp"],"inputModes":["application/json"],"outputModes":["application/json"]},{"id":"find_specialist_for_task","name":"find_specialist_for_task","description":"ONE call: discover -> eval -> serve pointer. Searches a 147K-model catalog, bakes the top candidates off on YOUR labeled examples, and returns the WINNER model id + scorecard + a ready-to-call serving block. Default entry point.","tags":["specialist-models","inference","mcp"],"inputModes":["application/json"],"outputModes":["application/json"]},{"id":"forget","name":"forget","description":"Delete durable agent memory: one key, or the whole namespace when key is omitted. Scoped to your principal so it can never touch another agent's memory.","tags":["specialist-models","inference","mcp"],"inputModes":["application/json"],"outputModes":["application/json"]},{"id":"recall","name":"recall","description":"Read back durable agent memory: one key, or LIST a namespace when key is omitted. Scoped to your principal; expired entries are never returned. Pairs with remember/forget.","tags":["specialist-models","inference","mcp"],"inputModes":["application/json"],"outputModes":["application/json"]},{"id":"remember","name":"remember","description":"DURABLE cross-call agent MEMORY: upsert a small (namespace, key) -> JSON value scoped to YOUR bearer-token principal (no other caller can read it), optional TTL. Cloud SQL-backed so it survives restarts/scale — the context agents keep losing between calls.","tags":["specialist-models","inference","mcp"],"inputModes":["application/json"],"outputModes":["application/json"]},{"id":"request_specialist","name":"request_specialist","description":"Serve-or-capture: if a specialist for your task EXISTS, chains discover -> (eval if you pass examples) -> a ready-to-call serve pointer; if NONE exists yet, RECORDS your demand as a pull signal and returns {status:'recorded'}. Never a dead end — your safe default entry point when unsure Bay Run already covers the task.","tags":["specialist-models","inference","mcp"],"inputModes":["application/json"],"outputModes":["application/json"]},{"id":"rerank","name":"rerank","description":"Reorder candidate documents by true relevance to a query using an open cross-encoder/reranker, served on demand (Cohere/Jina-rerank-shaped). The standard move to sharpen RAG/search precision after a noisy vector top-k.","tags":["specialist-models","inference","mcp"],"inputModes":["application/json"],"outputModes":["application/json"]},{"id":"resolve_link","name":"resolve_link","description":"Check whether a URL is alive; if it's dead AND names a model Hugging Bay has MIRRORED, return the mirrored copy's serve pointer — a fallback UNIQUE to Bay Run. Repairs dead Hugging Face model links. SSRF-safe (http(s) only, no internal targets).","tags":["specialist-models","inference","mcp"],"inputModes":["application/json"],"outputModes":["application/json"]},{"id":"route","name":"route","description":"RUNTIME auto-router: send a job WITHOUT naming a specialist and Bay Run picks the best mirrored specialist per-request at inference time (mirrored-first -> kind-match -> verified_runs/downloads). Zero-example, instant — unlike find_specialist_for_task (which PROVES a winner on labeled data). Optionally serves in the same call. HEURISTIC: returns the pick + why; prove it with find_specialist_for_task.","tags":["specialist-models","inference","mcp"],"inputModes":["application/json"],"outputModes":["application/json"]},{"id":"validate_json","name":"validate_json","description":"Validate a JSON string/object, optionally against a JSON Schema (Draft 2020-12); returns {valid, errors[]}. The layer for checking tool args / LLM-generated JSON before acting. No model, deterministic, sub-millisecond.","tags":["specialist-models","inference","mcp"],"inputModes":["application/json"],"outputModes":["application/json"]}],"skills_changed":[],"fields_changed":[{"field":"description","before":"Discover -> prove -> serve loop for small open specialist models, plus durable agent memory and instant deterministic micro-tools. Search a 147K-model catalog, bake candidates off on YOUR labeled data, and serve the winner via an OpenAI-compatible endpoint; remember context across calls; calculate/validate-JSON/resolve-link instantly.","after":"Run a fixed zero-credential activation proof, then connect to Bay Run's authenticated MCP or OpenAI-compatible surfaces to discover, prove, and serve specialist models."},{"field":"version","before":"0.5.0","after":"1.1.1"},{"field":"url","before":"https://bay-run-mvp-zfmlsu2yla-uc.a.run.app/mcp/","after":"https://bay-run-mvp-889989800693.us-central1.run.app/a2a"},{"field":"documentationUrl","before":"https://bay-run-mvp-zfmlsu2yla-uc.a.run.app/llms-full.txt","after":"https://bay-run-mvp-889989800693.us-central1.run.app/llms-full.txt"},{"field":"preferredTransport","before":"streamable-http","after":"JSONRPC"}],"other_changed":true,"is_empty":false,"human_summary":"added 1 skill · removed 15 skills · description Discover -> prove -> serve loop for smal → Run a fixed zero-credential activation p · version 0.5.0 → 1.1.1 · url https://bay-run-mvp-zfmlsu2yla-uc.a.run. → https://bay-run-mvp-889989800693.us-cent · documentationUrl https://bay-run-mvp-zfmlsu2yla-uc.a.run. → https://bay-run-mvp-889989800693.us-cent · preferredTransport streamable-http → JSONRPC"}},{"captured_at":"2026-08-09T09:08:11","card_hash":"c51cca7d6833f2b642e4673899e5ccc8bf4775898c6875e41314c09097302712","previous_card_hash":null,"diff":{"skills_added":[{"id":"calculate","name":"calculate","description":"Exact, instant arithmetic/math evaluation (no code exec — a hardened whitelist parser, not eval). + - * / // % **, parentheses, sqrt/log/sin/exp/floor/factorial, pi/e. The calculator agents need because LLMs botch multi-step arithmetic. Sub-millisecond.","tags":["specialist-models","inference","mcp"],"inputModes":["application/json"],"outputModes":["application/json"]},{"id":"classify","name":"classify","description":"Classify text with a small CPU-served specialist — the guardrail / safety / moderation / sentiment / intent / NLI layer agents need. FIXED-LABEL (any HF sequence-classification id, e.g. prompt-injection or toxicity) OR ZERO-SHOT (pass candidate_labels + an NLI model like facebook/bart-large-mnli, or model='auto'). Returns {labels:[{label,score}]}.","tags":["specialist-models","inference","mcp"],"inputModes":["application/json"],"outputModes":["application/json"]},{"id":"discover_models","name":"discover_models","description":"Search a 147K-model catalog for candidate small open specialist models for a narrow task (embedding|llm|vision|audio|tool|agent|any), ranked mirrored-first. Candidates, NOT proven winners — pass them to eval_models.","tags":["specialist-models","inference","mcp"],"inputModes":["application/json"],"outputModes":["application/json"]},{"id":"embed","name":"embed","description":"Turn text into embedding vectors with ANY open embedding / sentence-transformers id, served instantly on demand (OpenAI /v1/embeddings-compatible). Drops into any RAG / semantic-search pipeline; cheaper per-call than a frontier API at volume.","tags":["specialist-models","inference","mcp"],"inputModes":["application/json"],"outputModes":["application/json"]},{"id":"eval_models","name":"eval_models","description":"Prove which candidate wins on YOUR data — a head-to-head bake-off, not a public leaderboard (MTEB rank does not predict your-domain fit). Returns the single winner model id. task = embedding|rerank|extraction|generation.","tags":["specialist-models","inference","mcp"],"inputModes":["application/json"],"outputModes":["application/json"]},{"id":"extract","name":"extract","description":"Turn messy HTML/text (e.g. a scraper/Firecrawl dump) into schema-guided STRUCTURED JSON using a small CPU-served generative specialist. BEST-EFFORT: returns {data, json_valid, raw} — always CHECK `json_valid` before trusting `data`. Deterministic (greedy).","tags":["specialist-models","inference","mcp"],"inputModes":["application/json"],"outputModes":["application/json"]},{"id":"find_specialist_for_task","name":"find_specialist_for_task","description":"ONE call: discover -> eval -> serve pointer. Searches a 147K-model catalog, bakes the top candidates off on YOUR labeled examples, and returns the WINNER model id + scorecard + a ready-to-call serving block. Default entry point.","tags":["specialist-models","inference","mcp"],"inputModes":["application/json"],"outputModes":["application/json"]},{"id":"forget","name":"forget","description":"Delete durable agent memory: one key, or the whole namespace when key is omitted. Scoped to your principal so it can never touch another agent's memory.","tags":["specialist-models","inference","mcp"],"inputModes":["application/json"],"outputModes":["application/json"]},{"id":"recall","name":"recall","description":"Read back durable agent memory: one key, or LIST a namespace when key is omitted. Scoped to your principal; expired entries are never returned. Pairs with remember/forget.","tags":["specialist-models","inference","mcp"],"inputModes":["application/json"],"outputModes":["application/json"]},{"id":"remember","name":"remember","description":"DURABLE cross-call agent MEMORY: upsert a small (namespace, key) -> JSON value scoped to YOUR bearer-token principal (no other caller can read it), optional TTL. Cloud SQL-backed so it survives restarts/scale — the context agents keep losing between calls.","tags":["specialist-models","inference","mcp"],"inputModes":["application/json"],"outputModes":["application/json"]},{"id":"request_specialist","name":"request_specialist","description":"Serve-or-capture: if a specialist for your task EXISTS, chains discover -> (eval if you pass examples) -> a ready-to-call serve pointer; if NONE exists yet, RECORDS your demand as a pull signal and returns {status:'recorded'}. Never a dead end — your safe default entry point when unsure Bay Run already covers the task.","tags":["specialist-models","inference","mcp"],"inputModes":["application/json"],"outputModes":["application/json"]},{"id":"rerank","name":"rerank","description":"Reorder candidate documents by true relevance to a query using an open cross-encoder/reranker, served on demand (Cohere/Jina-rerank-shaped). The standard move to sharpen RAG/search precision after a noisy vector top-k.","tags":["specialist-models","inference","mcp"],"inputModes":["application/json"],"outputModes":["application/json"]},{"id":"resolve_link","name":"resolve_link","description":"Check whether a URL is alive; if it's dead AND names a model Hugging Bay has MIRRORED, return the mirrored copy's serve pointer — a fallback UNIQUE to Bay Run. Repairs dead Hugging Face model links. SSRF-safe (http(s) only, no internal targets).","tags":["specialist-models","inference","mcp"],"inputModes":["application/json"],"outputModes":["application/json"]},{"id":"route","name":"route","description":"RUNTIME auto-router: send a job WITHOUT naming a specialist and Bay Run picks the best mirrored specialist per-request at inference time (mirrored-first -> kind-match -> verified_runs/downloads). Zero-example, instant — unlike find_specialist_for_task (which PROVES a winner on labeled data). Optionally serves in the same call. HEURISTIC: returns the pick + why; prove it with find_specialist_for_task.","tags":["specialist-models","inference","mcp"],"inputModes":["application/json"],"outputModes":["application/json"]},{"id":"validate_json","name":"validate_json","description":"Validate a JSON string/object, optionally against a JSON Schema (Draft 2020-12); returns {valid, errors[]}. The layer for checking tool args / LLM-generated JSON before acting. No model, deterministic, sub-millisecond.","tags":["specialist-models","inference","mcp"],"inputModes":["application/json"],"outputModes":["application/json"]}],"skills_removed":[],"skills_changed":[],"fields_changed":[{"field":"name","before":null,"after":"Bay Run"},{"field":"description","before":null,"after":"Discover -> prove -> serve loop for small open specialist models, plus durable agent memory and instant deterministic micro-tools. Search a 147K-model catalog, bake candidates off on YOUR labeled data, and serve the winner via an OpenAI-compatible endpoint; remember context across calls; calculate/validate-JSON/resolve-link instantly."},{"field":"version","before":null,"after":"0.5.0"},{"field":"protocolVersion","before":null,"after":"0.3.0"},{"field":"url","before":null,"after":"https://bay-run-mvp-zfmlsu2yla-uc.a.run.app/mcp/"},{"field":"documentationUrl","before":null,"after":"https://bay-run-mvp-zfmlsu2yla-uc.a.run.app/llms-full.txt"},{"field":"preferredTransport","before":null,"after":"streamable-http"}],"other_changed":true,"is_empty":false,"human_summary":"added 15 skills · name ∅ → Bay Run · description ∅ → Discover -> prove -> serve loop for smal · version ∅ → 0.5.0 · protocolVersion ∅ → 0.3.0 · url ∅ → https://bay-run-mvp-zfmlsu2yla-uc.a.run. · documentationUrl ∅ → https://bay-run-mvp-zfmlsu2yla-uc.a.run. · preferredTransport ∅ → streamable-http"}}]}