Add AWF /reflect reference for gateway-based model discovery and routing

[Copilot]

New Feature

This PR adds a dedicated reference page for AWF GET /reflect, documenting how shared workflows/tools discover gateway inference endpoints, determine model availability, and choose models dynamically at runtime. It also codifies the requirement to route inference through the gateway rather than direct provider endpoints to preserve cost controls, tracking, and optimization signals.

• What does this feature do?

  • Adds docs/src/content/docs/reference/awf-reflect.md as the canonical /reflect documentation.
  • Documents response fields needed for runtime routing (provider, base_url, configured, models, models_url, models_fetch_complete).
  • Defines a practical selection flow: query /reflect → filter configured endpoints → prefer populated model lists → select provider/model dynamically.
  • Adds explicit gateway-only inference guidance for shared components.

• Why is this feature needed?

  • Shared workflows/tools run across environments with different provider configuration and model availability.
  • Hardcoded direct model endpoints break portability and bypass gateway-level governance/observability.

• Implementation details

  • Adds sidebar navigation entry in docs/astro.config.mjs under Reference (/reference/awf-reflect/).
  • Includes bounded retry guidance for transient models: null states during model discovery.

curl -s http://api-proxy:10000/reflect

Selection pattern:
1) GET /reflect
2) endpoints where configured == true
3) prefer endpoints with non-empty models
4) match requested alias/pattern against models
5) call selected endpoint.base_url (via gateway only)

[Copilot]

Pull request overview

Adds a new documentation reference page for AWF GET /reflect, intended to guide shared workflows/tools in discovering gateway inference endpoints, understanding model availability, and routing requests through the gateway.

Changes:

• Adds a new reference doc page: docs/src/content/docs/reference/awf-reflect.md.
• Documents recommended runtime selection flow and retry guidance for transient model discovery states.
• Adds a new sidebar entry under Reference in docs/astro.config.mjs.

Show a summary per file

File	Description
docs/src/content/docs/reference/awf-reflect.md	New /reflect reference page describing endpoint/model discovery and dynamic routing guidance.
docs/astro.config.mjs	Adds the new AWF Reflect Route page to the Reference sidebar navigation.

Copilot's findings

Tip

Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Comments suppressed due to low confidence (1)

docs/src/content/docs/reference/awf-reflect.md:43

• Step 6 suggests retrying discovery whenever models is null, but in practice models can be null for reasons that won’t change with retries (e.g., models_url is null, or model listing isn’t supported), and the more reliable signal for “still warming up” is models_fetch_complete plus whether models_url is present for a configured endpoint. Suggest adjusting the guidance to (a) only retry /reflect while models_fetch_complete is false, and/or (b) if configured is true and models_url is present, query models_url directly to populate models; otherwise treat models as unavailable.

1. Query `/reflect` at start of execution.
2. Filter endpoints to `configured: true`.
3. Prefer endpoints with a non-empty `models` list.
4. Match requested model aliases/patterns against available models.
5. Route inference to the selected endpoint `base_url`.
6. If `models` is `null`, retry discovery with bounded backoff (for example, every 3 seconds up to 5 attempts) before failing.

• Files reviewed: 2/2 changed files
• Comments generated: 1