Roadmap.

Model Catalog

Every model the API serves, with who serves it, its price per million tokens next to the provider's list price, its context window, capabilities and live availability.

Request-Level Execution Metadata

Every API response identifies its request and says which model served it, by which execution path, and what it cost, without exposing prompts or outputs.

Releases

1. Release 1

   Request IDs

   Every chat completion response carries its request id in x-avenro-request-id, errors included, and the dashboard's usage table shows each request's id.

   Phases 2, 8, 9

   Next
2. Release 2

   Model and exact cost

   Responses name the model that served them in x-avenro-model, and x-avenro-cost-usd states the charge only once it is recorded.

   Phases 3, 5

   Planned
3. Release 3

   Accurate token usage

   The usage in every response and stream reports exactly the tokens billed, reasoning included, whichever provider serves the model.

   Phases 4, 7

   Planned
4. Release 4

   Execution metadata guide

   A complete guide to the metadata in the docs, with Python, TypeScript and curl examples.

   Phase 10

   Planned

Phases

1. Phase 1

   Audit the existing implementation

   A review of the gateway, billing, logging and docs before any change, so the update builds on what already exists.

   Done · Oct 9, 2026
2. Phase 2

   Request ID generation

   A unique id for every request, generated by Avenro when the request arrives, sent in x-avenro-request-id and used for its usage record.

   Release 1 · Next
3. Phase 3

   Standardize response metadata

   One documented set of headers for the request id, the model, the execution path and the cost, with x-avenro-model added.

   Release 2 · Planned
4. Phase 4

   Token usage

   The usage object reports exactly the tokens billed, reasoning included, in OpenAI's format.

   Release 3 · Planned
5. Phase 5

   Exact request cost

   x-avenro-cost-usd states the charge only once it is recorded, always equal to what your balance is charged.

   Release 2 · Planned
6. Phase 6

   Execution-path metadata

   x-avenro-execution-path names the kind of deployment that served each request, after any failover. The API already sends it, so Update 02 keeps it as it is.

   Already live
7. Phase 7

   Streaming compatibility

   Streams keep their exact format, with the metadata in their headers and the tokens billed in the final usage chunk.

   Release 3 · Planned
8. Phase 8

   Error-response correlation

   Every error carries the request id, with status codes and error formats unchanged.

   Release 1 · Next
9. Phase 9

   Internal structured logging

   One structured log line per request for debugging, never with prompts, outputs or keys.

   Release 1 · Next
10. Phase 10

    Website and documentation changes

    A complete guide to the metadata, an updated API reference and SDK examples. The dashboard's request ID column comes with release 1.

    Release 4 · Planned