> For the complete documentation index, see [llms.txt](https://docs.trueopen.ai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.trueopen.ai/providers/model-integration/profile-manifest.md). # Profile Manifest Bind an adapter implementation to exact model Profiles, capabilities, artifacts, and runtime semantics. A capability manifest tells Cortex exactly what the adapter can execute. It is a technical claim about a concrete runtime, not a provider's price or willingness to accept work. ## Required identity Each advertised entry binds: * Model identifier and Profile version * Profile manifest digest * Runtime-capability digest * Inference, verification, or both capabilities * Required model, tokenizer, precision, and evidence artifacts * Runtime implementation and compatibility version The pair `(model_id, profile_version)` must be unique in one adapter view. Never treat a model name without its Profile version as sufficient identity. ## Capability semantics Inference support means the adapter can produce canonical output and every evidence object required by the Profile. Verification support means it can evaluate all committed generated positions and produce the required metric material. Advertising both means satisfying both contracts independently. Provider capacity, queue state, and minimum price are local operating policy and do not belong in the compatibility manifest. ## Readiness `ReadinessCheck` should verify artifacts, digests, runtime support, storage access, and a lightweight execution path for the exact Profile and capability. It must fail closed on missing or mismatched artifacts. A readiness response must identify what was checked and return the exact bindings expected by Cortex. Generic backend health is not Profile readiness. ## Version changes Any change that alters inference output semantics, verification metrics, evidence construction, or runtime compatibility requires the appropriate new Profile or capability digest. An in-place local change must not alter work already bound to an existing Task lease. Compute providers use these manifest results when they [configure Profiles and capabilities](/providers/profiles-and-capabilities.md). --- # Agent Instructions This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com. ## Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question. Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter: ``` GET https://docs.trueopen.ai/providers/model-integration/profile-manifest.md?ask=&goal= ``` `ask` is the immediate question: it should be specific, self-contained, and written in natural language. `goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.