feat: mature gateway webui api surfaces

[chumyin]

Summary

• Problem:
  Loong's gateway already exposed useful HTTP surfaces, but third-party WebUI integration was still incomplete. Operators could talk to /v1/chat/completions, /v1/models, and /v1/turn, yet the gateway still lacked a persisted Responses-style continuity surface and had rough streaming-compatibility edges.
• Why it matters:
  Operators want to connect Loong to their own browser frontends or OpenAI-compatible chat UIs without building a custom bridge around the runtime first.
• What changed:
  Added a persisted gateway Responses surface, completed fetch/delete response endpoints, taught chat-completions streaming to degrade gracefully on non-streaming providers, wired request-level tool-view overrides into turn execution, and documented the trusted WebUI contract.
• What did not change (scope boundary):
  Caller-supplied tools remain limited to Loong's structured gateway tool layer. Direct runtime tools are still governed by Loong's own runtime surface rather than arbitrary client-declared tool contracts.

Linked Issues

• Closes Mature gateway WebUI API surfaces for third-party frontends #1438
• Related #

Change Type

• Bug fix
• Feature
• Refactor
• Documentation
• Security hardening
• CI / workflow / release

Touched Areas

• Kernel / policy / approvals
• Contracts / protocol / spec
• Daemon / CLI / install
• Providers / routing
• Tools
• Browser automation
• Channels / integrations
• ACP / conversation / session runtime
• Memory / context assembly
• Config / migration / onboarding
• Docs / contributor workflow
• CI / release / workflows

Risk Track

• Track A (routine / low-risk)
• Track B (higher-risk / policy-impacting)

If Track B, fill these in:

• Risk notes:
  This changes the public gateway contract for WebUI clients and threads request-level tool settings into the turn-runtime seam.
• Rollout / guardrails:
  The new contract is covered by focused gateway tests and keeps direct runtime tools behind Loong-controlled governance instead of opening arbitrary external tool contracts.
• Rollback path:
  Revert this branch or disable use of /v1/responses and request-level tool fields on the client side; existing /v1/turn remains available as the native control surface.

Validation

• cargo fmt --all -- --check
• cargo clippy --workspace --all-targets --all-features -- -D warnings
• cargo test --workspace --locked
• cargo test --workspace --all-features --locked
• Relevant architecture / dep-graph / docs checks for touched areas
• Additional scenario, benchmark, or manual checks when behavior changed
• If this changes config/env fallback, limits, or defaults: include before/after behavior and regression coverage for explicit path, fallback path, and boundary values
• If tests mutate process-global env: document how state is restored or serialized

Commands and evidence:

./scripts/cargo-local-toolchain.sh fmt --all -- --check
./scripts/cargo-local-toolchain.sh test -p loong gateway_ --lib
./scripts/cargo-local-toolchain.sh clippy -p loong --all-targets --all-features -- -D warnings

Result:
- fmt check passed
- gateway_ test surface passed (42 tests)
- package-scoped clippy passed with -D warnings

User-visible / Operator-visible Changes

• Trusted WebUI clients can now use /v1/responses with persisted response continuity.
• Clients can fetch and delete stored responses through the gateway.
• chat/completions streaming now degrades gracefully when the configured provider can complete turns but does not expose native token streaming.
• tool_choice="none" is accepted on the gateway-compatible request surface.
• Public docs now include a dedicated WebUI gateway integration page.

Failure Recovery

• Fast rollback or disable path:
  Revert the commit series or stop advertising /v1/responses to clients and continue using /v1/chat/completions or /v1/turn.
• Observable failure symptoms reviewers should watch for:
  Missing response persistence, continued responses losing history, gateway tool filters dropping the expected structured gateway tools, or streaming responses falling back to empty SSE bodies.

Reviewer Focus

• Review the request-level tool-view override seam in crates/app/src/agent_runtime.rs, crates/app/src/chat.rs, and crates/app/src/conversation/runtime.rs.
• Review the gateway contract additions in crates/daemon/src/gateway/openai_compat.rs, openai_responses.rs, openai_tooling.rs, and response_store.rs.
• Review the public contract wording in site/use-loong/gateway-webui.mdx to confirm the stated boundary around structured gateway tools vs direct runtime tools is the one we want to support.