diff --git a/WORK-RECORDS.md b/WORK-RECORDS.md index 2338d79..382808e 100644 --- a/WORK-RECORDS.md +++ b/WORK-RECORDS.md @@ -17,5 +17,5 @@ | task | GLAS-WP-0001-T02 | done | — | workplans/GLAS-WP-0001-harness-router-foundation.md | | task | GLAS-WP-0001-T03 | done | — | workplans/GLAS-WP-0001-harness-router-foundation.md | | task | GLAS-WP-0001-T04 | progress | — | workplans/GLAS-WP-0001-harness-router-foundation.md | -| task | GLAS-WP-0001-T05 | progress | — | workplans/GLAS-WP-0001-harness-router-foundation.md | +| task | GLAS-WP-0001-T05 | done | — | workplans/GLAS-WP-0001-harness-router-foundation.md | | task | GLAS-WP-0001-T06 | done | — | workplans/GLAS-WP-0001-harness-router-foundation.md | diff --git a/workplans/GLAS-WP-0001-harness-router-foundation.md b/workplans/GLAS-WP-0001-harness-router-foundation.md index d710a04..26e89a0 100644 --- a/workplans/GLAS-WP-0001-harness-router-foundation.md +++ b/workplans/GLAS-WP-0001-harness-router-foundation.md @@ -76,12 +76,24 @@ Wiring is built and unit-tested: `src/glas_harness/gateway.py` + `reins/rein_aharness.py` implement start_session/dispatch_tool/ end_session, tested against a real local git repo and a mocked `SandboxManager`/`agent-harness` subprocess (`tests/test_gateway.py`, -`tests/test_rein_aharness.py`). **Not yet done:** an actual live run -(`glas-harness run --sandbox-profile profile.bwrap-local ...`) against -the real Claude Code CLI — that spends real API credits/credentials and -was deliberately left for a human-triggered run rather than executed -autonomously. State Hub reporting from the gateway itself is also not -wired yet (only rein-aharness's own `--no-hub` path was exercised). +`tests/test_rein_aharness.py`). + +**The gateway pattern is now live-proven — via `rein-openweights`, not +yet `rein-aharness`.** With the user's explicit go-ahead, ran +`run_task_through_rein(sandbox_profile="profile.bwrap-local", ..., +rein=ReinOpenWeights(model="qwen/qwen-2.5-72b-instruct"))` for real: a +real `ext.bwrap` sandbox was provisioned, a real OpenRouter call drove a +2-turn tool-calling loop, a real commit landed +(`2effe57ac62ab7a5e1cd59caa86d4c9c3338bf10`) inside the ephemeral +sandbox workspace, verified via `git rev-parse HEAD`, and the sandbox +was torn down — `executor-sandbox` itself was untouched (bwrap only +mutates its copy). Full result: `tool_ok: true`, `committed: "True"`. +This proves the glas-harness → sand-boxer → rein → verified-commit path +end-to-end for one rein. **Still not done:** the same live proof against +the real Claude Code CLI via `rein-aharness` specifically — that spends +different credentials and wasn't part of this run. State Hub reporting +from the gateway itself also remains unwired (`--no-hub` was passed +through to the rein's own CLI both times). ```task id: GLAS-WP-0001-T04