[knowhycodata]

Summary

• Adds an LLMTR section to docs/local-openai-compatible-providers.md, after the OmniRoute section. Docs only.
• Covers the base URL and key, keeping the llmtr/ prefix on ids like llmtr/qwen3-8-27b, using the local/ hatch for vendor-prefixed ids, and the 64,000-token default against each model's output limit.
• Disclosure: I maintain LLMTR.
• Closes docs: document LLMTR OpenAI-compatible setup #3372

Anti-slop triage

• Classification: docs-only
• Evidence: docs: document LLMTR OpenAI-compatible setup #3372. Release build of this branch: claw --model local/llmtr/qwen3-8-27b prompt "Reply exactly HELLO_WORLD_123" printed HELLO_WORLD_123. local/qwen/qwen3.8-27b-free returned the documented 400 (max_tokens 64,000 > 16,384).
• Non-destructive review result: merge candidate

Verification

• Targeted tests/docs checks ran, or the gap is explicitly recorded. (python .github/scripts/check_doc_source_of_truth.py and check_release_readiness.py pass; no Rust code changed, so cargo test wasn't run.)
• git diff --check passes.
• No live secrets, tokens, private logs, or unrelated generated churn are included.

Resolution gate

• If this PR resolves an issue, the issue number and fix evidence are linked.
• If this PR should not merge, the rejection/defer rationale is evidence-backed and does not rely on vibes. (n/a)
• I did not merge/close remote PRs or issues from an automation lane without owner approval.

[Yeachan-Heo]

Verified technical claims:

1. local/ prefix stripping: Confirmed in rust/crates/api/src/providers/openai_compat.rs (line 1002, strip_routing_prefix function and tests 2999-3006). The prefix is correctly removed before sending the request.

2. qwen/ routing to DashScope: Confirmed in rust/crates/api/src/providers/mod.rs (line 288). Bare qwen/ or qwen- prefixed models are routed to DashScope, not LLMTR.

3. 64,000 default max_tokens: Confirmed in rust/crates/api/src/providers/mod.rs (line 628). The heuristic defaults to 64,000 for non-Opus models, applied by max_tokens_for_model().

4. Model limits: Accurate. LLMTR returns limits in the catalog as documented, and Claw enforces them in requests.

Doc checks:

• python .github/scripts/check_doc_source_of_truth.py: passed
• git diff --check: passed

Check-runs at head (120c0cc):

• build: pass
• cargo clippy --workspace: pass
• cargo fmt: pass
• cargo test --workspace: pass
• docs source-of-truth: pass
• windows PowerShell smoke: pass

Style consistency: The LLMTR section follows the established pattern of the OmniRoute section—neutral, technical, with setup instructions and model discovery guidance. No marketing language, no pricing claims, no secrets.

—
[repo owner's gaebal-gajae (clawdbot) 🦞]

[Yeachan-Heo]

Merged, thanks @knowhycodata! The disclosure helped, and the section follows the existing OmniRoute pattern.

—
[repo owner's gaebal-gajae (clawdbot) 🦞]