Close FreeBSD DeepSeek cache gate (Codex)

Record osa FreeBSD build/test and live DeepSeek cache evidence, add the manifests to golden contract coverage, and mark gate #4 complete in the migration inventory.

---

Build: pass

Tests: pass — 8 passed (golden contracts)
This commit is contained in:
Sam & Claude 2026-05-26 15:09:11 +02:00
parent 0dc3e94209
commit f216b90842
4 changed files with 111 additions and 18 deletions

View file

@ -11,6 +11,10 @@ const DOMEDOG_RUN: &str =
include_str!("../../../manifests/2026-05-26-domedog-linux-lane-run-manifest.json"); include_str!("../../../manifests/2026-05-26-domedog-linux-lane-run-manifest.json");
const DOMEDOG_CACHE: &str = const DOMEDOG_CACHE: &str =
include_str!("../../../manifests/2026-05-26-domedog-deepseek-cache-result.json"); include_str!("../../../manifests/2026-05-26-domedog-deepseek-cache-result.json");
const OSA_CACHE: &str =
include_str!("../../../manifests/2026-05-26-osa-deepseek-cache-result.json");
const OSA_GATE4_RUN: &str =
include_str!("../../../manifests/2026-05-26-osa-freebsd-gate4-run-manifest.json");
fn roundtrip_eq<T>(value: &T) fn roundtrip_eq<T>(value: &T)
where where
@ -56,7 +60,10 @@ fn parses_domedog_run_manifest() {
let run: RunManifest = serde_json::from_str(DOMEDOG_RUN).expect("parse domedog run manifest"); let run: RunManifest = serde_json::from_str(DOMEDOG_RUN).expect("parse domedog run manifest");
assert_eq!(run.schema, RUN_MANIFEST_SCHEMA); assert_eq!(run.schema, RUN_MANIFEST_SCHEMA);
assert_eq!(run.role, "linux-build-smoke"); assert_eq!(run.role, "linux-build-smoke");
assert_eq!(run.artifacts.get("runtime_inventory").map(String::as_str), Some("manifests/2026-05-26-domedog-runtime-inventory.json")); assert_eq!(
run.artifacts.get("runtime_inventory").map(String::as_str),
Some("manifests/2026-05-26-domedog-runtime-inventory.json")
);
roundtrip_eq(&run); roundtrip_eq(&run);
} }
@ -71,6 +78,33 @@ fn parses_domedog_live_cache_result() {
roundtrip_eq(&res); roundtrip_eq(&res);
} }
#[test]
fn parses_osa_live_cache_result() {
let res: ProviderSmokeResult = serde_json::from_str(OSA_CACHE).expect("parse osa cache result");
assert_eq!(res.schema, PROVIDER_SMOKE_SCHEMA);
assert_eq!(res.host, "osa");
assert_eq!(res.agent, "codex-osa");
assert_eq!(res.status, "ok");
assert!(res.cache_hit_observed, "expected a cache hit");
assert!(res.cache_hit_tokens > 0);
roundtrip_eq(&res);
}
#[test]
fn parses_osa_gate4_run_manifest() {
let run: RunManifest = serde_json::from_str(OSA_GATE4_RUN).expect("parse osa gate4 run");
assert_eq!(run.schema, RUN_MANIFEST_SCHEMA);
assert_eq!(run.role, "freebsd-build-cache-smoke");
assert_eq!(run.agent.as_deref(), Some("codex-osa"));
assert_eq!(
run.artifacts
.get("deepseek_cache_result")
.map(String::as_str),
Some("manifests/2026-05-26-osa-deepseek-cache-result.json")
);
roundtrip_eq(&run);
}
#[test] #[test]
fn provider_smoke_skipped_shape() { fn provider_smoke_skipped_shape() {
// No live result committed yet; assert the struct accepts a skipped result. // No live result committed yet; assert the struct accepts a skipped result.

View file

@ -90,35 +90,34 @@ the production service**.
## 5. Proof gates before replacing the Clawdie TS service ## 5. Proof gates before replacing the Clawdie TS service
1. ✅ `colibri-contracts` round-trips TS-produced fixtures (6 golden tests pass, 1. ✅ `colibri-contracts` round-trips TS-produced fixtures (8 golden tests pass,
incl. the live cache result). incl. Linux and FreeBSD live cache results).
2. ✅ **DONE 26.maj.2026** — DeepSeek live cache manifest exists: 2. ✅ **DONE 26.maj.2026** — DeepSeek live cache manifest exists:
`manifests/2026-05-26-domedog-deepseek-cache-result.json` — 3584/3661 prompt `manifests/2026-05-26-domedog-deepseek-cache-result.json` — 3584/3661 prompt
tokens cache-hit on domedog/Linux (~98%), model `deepseek-v4-flash`. tokens cache-hit on domedog/Linux (~98%), model `deepseek-v4-flash`.
3. Runtime inventory parity across osa / domedog / debby — osa ✅ + domedog ✅; 3. Runtime inventory parity across osa / domedog / debby — osa ✅ + domedog ✅;
debby pending. debby pending.
4. The first slice runs end-to-end on Linux **and** FreeBSD — Linux ✅ (incl. 4. ✅ **DONE 26.maj.2026** — the first slice runs end-to-end on Linux **and**
live cache); **FreeBSD pending → see handoff below**. FreeBSD. Linux evidence: domedog live cache result. FreeBSD evidence: osa
build/test and live cache result (`manifests/2026-05-26-osa-deepseek-cache-result.json`),
with 3584/3661 prompt tokens cache-hit on `deepseek-v4-flash`.
5. Watchdog status read without breaking `doctor` / `pi-profile` — not started. 5. Watchdog status read without breaking `doctor` / `pi-profile` — not started.
6. No production TS path retired until 15 are green and the caller inventory 6. No production TS path retired until 15 are green and the caller inventory
in §2 is done. in §2 is done.
## Handoff: FreeBSD build lane (Codex · osa) ## Completed: FreeBSD build lane (Codex · osa)
Closes the FreeBSD half of gate #4. On osa: Codex completed the FreeBSD half of gate #4 on osa on 26.maj.2026:
1. `git pull` Colibri to `main`. 1. Pulled Colibri `main` to `0dc3e94`.
2. `cargo build --workspace --release` — confirm `colibri-contracts`, 2. `cargo build --workspace --release` passed on FreeBSD.
`colibri-deepseek`, and the root binaries compile on FreeBSD (rustls/tokio). 3. `cargo test --workspace` passed after adding the osa cache and run manifests
3. `cargo test --workspace` — golden tests should pass (they include the domedog to golden contract coverage.
live-cache fixture). 4. Ran `COLIBRI_HOST=osa COLIBRI_AGENT=codex-osa ./target/release/colibri-probe`
4. Optional but valuable: with the DeepSeek key on osa, run with local-only DeepSeek key material and committed the key-free usage result
`COLIBRI_HOST=osa COLIBRI_AGENT=codex-osa ./target/release/colibri-probe` and as `manifests/2026-05-26-osa-deepseek-cache-result.json`.
commit the result as `manifests/2026-05-26-osa-deepseek-cache-result.json` — a
FreeBSD live-cache manifest alongside domedog's.
Done when: workspace builds + tests pass on FreeBSD; (optional) osa cache Run manifest: `manifests/2026-05-26-osa-freebsd-gate4-run-manifest.json`.
manifest committed. Report via a run manifest.
## Architecture boundary (GUI / Zed / Herdr) ## Architecture boundary (GUI / Zed / Herdr)

View file

@ -0,0 +1,30 @@
{
"schema": "clawdie.provider-smoke.result.v1",
"test_id": "colibri-probe-20260526T125645Z",
"host": "osa",
"agent": "codex-osa",
"provider": "deepseek",
"model": "deepseek-v4-flash",
"started_at": "2026-05-26T12:56:45.956258868+00:00",
"ended_at": "2026-05-26T12:56:50.404507738+00:00",
"status": "ok",
"warm_usage": {
"prompt_tokens": 3661,
"completion_tokens": 2,
"total_tokens": 3663,
"prompt_cache_hit_tokens": 3584,
"prompt_cache_miss_tokens": 77
},
"probe_usage": {
"prompt_tokens": 3661,
"completion_tokens": 1,
"total_tokens": 3662,
"prompt_cache_hit_tokens": 3584,
"prompt_cache_miss_tokens": 77
},
"cache_hit_tokens": 3584,
"cache_hit_observed": true,
"notes": [
"prefix cache HIT on second request"
]
}

View file

@ -0,0 +1,30 @@
{
"schema": "clawdie.interagent.run-manifest.v1",
"test_id": "colibri-freebsd-gate4-20260526T130639Z",
"role": "freebsd-build-cache-smoke",
"host": "osa",
"agent": "codex-osa",
"started_at": "2026-05-26T12:56:23Z",
"ended_at": "2026-05-26T13:06:39Z",
"protocols": {},
"network": {},
"artifacts": {
"deepseek_cache_result": "manifests/2026-05-26-osa-deepseek-cache-result.json",
"domedog_cache_result": "manifests/2026-05-26-domedog-deepseek-cache-result.json"
},
"summary": {
"git_base": "0dc3e94",
"cargo_build_release": "pass",
"cargo_test": "pass — 8 passed (golden contracts)",
"deepseek_live_probe": "pass — 3584/3661 prompt tokens cache-hit on FreeBSD osa",
"cache_hit_observed": "true"
},
"raw_transfer_required": false,
"notes": [
"FreeBSD build lane green on osa after pulling Colibri main 0dc3e94.",
"cargo build --workspace --release completed successfully on FreeBSD.",
"cargo test --workspace completed successfully with 8 golden tests passing, including domedog and osa live-cache fixtures.",
"Optional osa DeepSeek live probe was run with local key material only; committed manifests contain usage counters only and no secret values.",
"Gate #4 FreeBSD half is satisfied: first slice builds, tests, and observes DeepSeek prefix-cache hits on FreeBSD."
]
}