coilyco-flight-deck/agent-proxy

Watch

[proxy 07/13] Live num_ctx proof script against the tower #11

New issue

Closed

opened 2026-06-26 09:48:19 +00:00 by coilysiren · 2 comments

coilysiren commented

2026-06-26 09:48:19 +00:00

Owner

Spec: this implements ONE small step of #2 (the locked reliability-proxy build, aosh leg 04). Do not redesign. Stay tightly scoped to this single step.

Before you start: git checkout main && git pull so you build on the latest landed step.
Add scripts/prove_num_ctx.py - the end-to-end proof that the 32k cliff is gone.

Resolve the tower FQDN from SSM /coilysiren/kai-tower-3026/tailnet-fqdn (or env override TOWER_FQDN), backend http://$TOWER:11434.
Build a ~55k-token fast-think chat request and send it THROUGH the proxy (POST http://127.0.0.1:8080/v1/chat/completions).
Assert usage.prompt_tokens (from the upstream prompt_eval_count) is well above 32767 (near 49152), printing PASS/FAIL.
This live run is gated on the tailnet route ward#330. If the tower is unreachable, the script must print SKIPPED (tower unreachable, see ward#330) and exit 0 - do NOT fail the build on it.

Acceptance: the script exists, runs, and either PASSes against the tower or cleanly SKIPs.

When done:

pre-commit run --all-files (or the repo's gate) must pass, commit your work, and push to canonical main (this is an ephemeral feature container - finish the merge, do not leave it for a human).
Update the matching line in docs/FEATURES.md from planned to landed for this feature, in the same push.
Dispatch the next step: ward agent engineer coilyco-flight-deck/agent-proxy#12 --driver goose

**Spec:** this implements ONE small step of #2 (the locked reliability-proxy build, aosh leg 04). **Do not redesign. Stay tightly scoped to this single step.** **Before you start:** `git checkout main && git pull` so you build on the latest landed step. Add `scripts/prove_num_ctx.py` - the end-to-end proof that the 32k cliff is gone. - Resolve the tower FQDN from SSM `/coilysiren/kai-tower-3026/tailnet-fqdn` (or env override `TOWER_FQDN`), backend `http://$TOWER:11434`. - Build a ~55k-token `fast-think` chat request and send it THROUGH the proxy (`POST http://127.0.0.1:8080/v1/chat/completions`). - Assert `usage.prompt_tokens` (from the upstream `prompt_eval_count`) is well above 32767 (near 49152), printing PASS/FAIL. - **This live run is gated on the tailnet route ward#330.** If the tower is unreachable, the script must print `SKIPPED (tower unreachable, see ward#330)` and exit 0 - do NOT fail the build on it. **Acceptance:** the script exists, runs, and either PASSes against the tower or cleanly SKIPs. **When done:** 1. `pre-commit run --all-files` (or the repo's gate) must pass, commit your work, and push to canonical `main` (this is an ephemeral feature container - finish the merge, do not leave it for a human). 2. Update the matching line in `docs/FEATURES.md` from `planned` to `landed` for this feature, in the same push. 3. Dispatch the next step: `ward agent engineer coilyco-flight-deck/agent-proxy#12 --driver goose`

coilyco-ops added the

autonomy

async-consult

priority

labels

2026-07-02 16:42:40 +00:00

coilysiren commented

2026-07-02 16:48:40 +00:00

Author

Owner

🔒 Reserved by ward agent --driver goose — container engineer-goose-agent-proxy-11 on host 7d608a8a9f0e is carrying this issue (reserved 2026-07-02T16:48:40Z). Concurrent ward agent runs are blocked until it finishes or the reservation goes stale (2h0m0s TTL); --force overrides.

— Goose, via ward agent

🔒 Reserved by `ward agent --driver goose` — container `engineer-goose-agent-proxy-11` on host `7d608a8a9f0e` is carrying this issue (reserved 2026-07-02T16:48:40Z). Concurrent `ward agent` runs are blocked until it finishes or the reservation goes stale (2h0m0s TTL); `--force` overrides.  — Goose, via `ward agent`

coilysiren referenced this issue

2026-07-02 16:49:18 +00:00

[proxy 06/13] /v1/chat/completions (non-streaming) shaped to OpenAI schema #10

coilysiren commented

2026-07-03 19:33:30 +00:00

Author

Owner

Landed. The live num_ctx proof shipped as scripts/truncation_proof.py (present on main), exposed as ward exec proof, and FEATURES.md records the end-to-end proof. Closing as done. Live-tower nuance (ollama's OLLAMA_NUM_PARALLEL halving the effective num_ctx) is tracked separately in #33.

Landed. The live num_ctx proof shipped as scripts/truncation_proof.py (present on main), exposed as `ward exec proof`, and FEATURES.md records the end-to-end proof. Closing as done. Live-tower nuance (ollama's OLLAMA_NUM_PARALLEL halving the effective num_ctx) is tracked separately in #33.

coilysiren closed this issue

2026-07-03 19:33:31 +00:00

No milestone

No project

No assignees

1 participant

Notifications

Due date

The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set

Reference

coilyco-flight-deck/agent-proxy#11

No description provided.

Rows
Columns