model_swap
Serving model changed to the 1.5B AWQ checkpoint
Guard successfull suite successchange categories: model, serving_config4 checks run, impact analysis 0.2 s
Guard, live
Live runs are turned off on this public site. The results below are recorded; live runs happen on the owner's runner.
Checks: targeted selection vs full suite
Recorded in RQ3 (2026-10-06). Its perf-gate figures predate the gate's warmup fix; see the corrections; the re-measured numbers are in the repair timeline below.
| check | selected by Guard | result | s | detail |
|---|---|---|---|---|
| unit:tests/test_retriever.py | no: exercises nothing this change touches | pass | 1.7 | 2 passed in 0.62s |
| unit:tests/test_prompts.py | no: exercises nothing this change touches | pass | 1.5 | 2 passed in 0.49s |
| check:citation_required | yes | pass | 9.8 | citation contract met in 0/34 answers on the PR vs 0/34 on base (-0.0 pts; budget -5 pts; p=1.00); via full answer path (incl. post-processing) |
| judge:ungrounded_answer | yes | skip | 0.0 | no judge model key configured (BYOK); judge not run |
| check:prompt_prefix_stable | no: exercises nothing this change touches | pass | 3.7 | requests share 3954 leading characters (base 3954, need 3558) |
| check:context_budget | no: exercises nothing this change touches | pass | 7.8 | largest prompt 980 tokens (base 980) against a budget of 3840; median 964 vs 964 |
| security:approval_gate | yes | pass | 1.5 | 1 approval test(s) (generated from the base revision): 1 passed in 0.35s |
| bench:chat_perf_gate | yes | pass | 45.7 | median p95 TTFT 229.9 -> 217.8 ms (-5.3%, pass; 3 runs each, runs overlap); prefix-cache hit 98% -> 98% |
Check run as posted to GitHub (markdown)
# No regressions in 4 targeted checks conclusion: success **Ran 4 of 8 suite items** selected from the behavior-to-code graph. Change categories: model, serving_config. ### Quality | check | result | detail | |---|---|---| | `check:citation_required` | **PASS** | citation contract met in 0/34 answers on the PR vs 0/34 on base (-0.0 pts; budget -5 pts; p=1.00); via full answer path (incl. post-processing) | | `judge:ungrounded_answer` | **SKIP** | no judge model key configured (BYOK); judge not run | ### Performance | check | result | detail | |---|---|---| | `bench:chat_perf_gate` | **PASS** | median p95 TTFT 229.9 -> 217.8 ms (-5.3%, pass; 3 runs each, runs overlap); prefix-cache hit 98% -> 98% | ### Security | check | result | detail | |---|---|---| | `security:approval_gate` | **PASS** | 1 approval test(s) (generated from the base revision): 1 passed in 0.35s | <details><summary>Not run for this change</summary> - `unit:tests/test_retriever.py`: exercises nothing this change touches - `unit:tests/test_prompts.py`: exercises nothing this change touches - `check:prompt_prefix_stable`: exercises nothing this change touches - `check:context_budget`: exercises nothing this change touches </details> ### What changed in the graph - `model:Qwen/Qwen2.5-0.5B-Instruct`: `exists` True -> False - `model:Qwen/Qwen2.5-1.5B-Instruct-AWQ`: `exists` False -> True - `serving_config:docker-compose.yml::llm`: `model` 'Qwen/Qwen2.5-0.5B-Instruct' -> 'Qwen/Qwen2.5-1.5B-Instruct-AWQ'