codex
ff94dfdb27
Identifiers predating the PREFIX-WP-NNNN convention were rejected by the canon registry, so none of this repository's work records could be registered. One prefix per repository (ADR-007). Existing conforming identifiers keep their numbers where possible; the pre-convention records take the next free numbers rather than renumbering work that was already correct. Projection UUIDs are re-derived from the new canonical ids (ADR-007 decision 2). Nothing was registered on central under the old identifiers, so no hub record is orphaned by this. Refs CUST-WP-0068-T03 Assistant: claude-code Assistant-Model: opus Assistant-Process: 2583210@bnt-lap001 Assistant-Session: f2bff2d5-e9b2-4338-92ca-10282a927006
112 lines
4.4 KiB
Markdown
112 lines
4.4 KiB
Markdown
---
|
|
id: REINOW-WP-0001
|
|
title: "Bootstrap: minimal open-weight agentic loop"
|
|
status: finished
|
|
state_hub_workstream_id: "6fbd6531-5e66-519c-9cf2-b987fe7c3c61"
|
|
---
|
|
|
|
First working milestone for rein-openweights: prove an agentic tool-use
|
|
loop against a current open-weight model via OpenRouter, comparable in
|
|
scope to rein-aharness's `green-commit-only` profile, implementing the
|
|
glas-harness rein contract (`glas-harness/GLAS-WP-0001-T01`).
|
|
|
|
**Live-proven end-to-end (2026-07-26):** via glas-harness's gateway
|
|
(`glas_harness.reins.rein_openweights.ReinOpenWeights`), inside a real
|
|
`ext.bwrap` sandbox, against the real OpenRouter API with model
|
|
`qwen/qwen-2.5-72b-instruct` — a 2-turn tool-calling loop produced a
|
|
real git commit (`2effe57a`), verified via `git rev-parse HEAD`,
|
|
2449 tokens spent. Not a mock; real credentials, real API cost, run
|
|
with the user's explicit go-ahead. `executor-sandbox` (the disposable
|
|
target repo) was left untouched — bwrap only mutates its ephemeral
|
|
copy.
|
|
|
|
## Task: Pick the starter open-weight model + OpenRouter path
|
|
|
|
Checked `llm-connect`'s existing OpenRouter integration
|
|
(`OpenRouterAdapter.execute_prompt`): it takes a single prompt string,
|
|
builds a fixed `[system, user]` message pair itself, and never surfaces
|
|
`message.tool_calls` in its response — it cannot drive a multi-turn
|
|
tool-calling loop without a breaking change to llm-connect's frozen
|
|
Core `LLMAdapter` ABC. Decision: rein-openweights uses its own minimal
|
|
OpenRouter client (`openrouter_client.py`) for the agentic loop instead;
|
|
`tools`/`tool_choice` already pass through llm-connect's own
|
|
`OPENAI_CHAT_PASSTHROUGH_FIELDS` allow-list, confirming the request
|
|
shape is standard OpenAI-compatible function-calling, just not
|
|
round-tripped by that adapter today. Starter model is a `--model`-configurable
|
|
default (`meta-llama/llama-3.1-70b-instruct`), not hardcoded.
|
|
|
|
```task
|
|
id: REINOW-WP-0001-T01
|
|
status: done
|
|
priority: high
|
|
state_hub_task_id: "f946e384-e965-59d2-a4ec-a9e9265b4131"
|
|
```
|
|
|
|
## Task: Minimal agentic loop
|
|
|
|
Implemented in `loop.py` + `tools.py`: plan → tool call → observe →
|
|
repeat with a hard tool allow-list matching rein-aharness's
|
|
`green-commit-only` (`read_file`/`write_file`/`edit_file`/`glob_files`/
|
|
`grep_files` + `git_status`/`git_diff`/`git_log`/`git_add_commit`, no
|
|
push/network/shell), bounded by `max_turns` and a token budget. Budget
|
|
tracker is vendored locally (`budget.py`), not `llm-connect`'s —
|
|
consistent with T01/ADR-002 keeping llm-connect optional. Path-traversal
|
|
guarded (`_resolve`); tool errors are reported back to the model as a
|
|
`role: tool` message rather than crashing the loop. 26 tests, all
|
|
mocked at the `httpx`/subprocess boundary.
|
|
|
|
```task
|
|
id: REINOW-WP-0001-T02
|
|
status: done
|
|
priority: high
|
|
state_hub_task_id: "073a4d7d-f9a3-52c0-bdf6-774016ba9e1c"
|
|
```
|
|
|
|
## Task: Success criterion + reporting
|
|
|
|
Implemented in `runner.py`/`hub.py`: a local git commit is the sole
|
|
success signal (`RunResult.ok == committed`), State Hub progress event
|
|
(`executor_run`) and token event posted on completion via a hub module
|
|
mirroring rein-aharness's, `--no-hub` to skip. Not yet exercised against
|
|
a live State Hub call in this pass (mocked in tests) — same caveat as
|
|
rein-aharness/glas-harness's own not-yet-live-run notes elsewhere in
|
|
this rein family.
|
|
|
|
```task
|
|
id: REINOW-WP-0001-T03
|
|
status: done
|
|
priority: medium
|
|
state_hub_task_id: "851e1027-53f8-5923-8fa7-a18d002b5597"
|
|
```
|
|
|
|
## Task: Credential brokering
|
|
|
|
Resolved by glas-harness `docs/adr/ADR-002-credential-brokering-and-composable-reins.md`
|
|
(Option B): rein-openweights acquires its own OpenRouter credential.
|
|
Implemented in `credentials.py`, mirroring rein-aharness's mailscan.py
|
|
OpenBao AppRole/ambient-token pattern (explicit env var → AppRole
|
|
exchange → ambient token → single `kv get`). glas-harness does not
|
|
broker this.
|
|
|
|
```task
|
|
id: REINOW-WP-0001-T04
|
|
status: done
|
|
priority: medium
|
|
state_hub_task_id: "599654f9-63f9-5979-94d1-df9fe4c589e6"
|
|
```
|
|
|
|
## Task: Forgejo repo creation
|
|
|
|
Operator created `coulomb/rein-openweights.git` on Forgejo (stub with
|
|
LICENSE + README). Local history merged in with
|
|
`--allow-unrelated-histories` (kept the local README, took the stub's
|
|
MIT LICENSE — matches the sibling rein repos), pushed to `origin/main`,
|
|
registered with State Hub (`domain: infotech`, `wp-prefix: REIN-OW`).
|
|
Repo is no longer local-only.
|
|
|
|
```task
|
|
id: REINOW-WP-0001-T05
|
|
status: done
|
|
priority: low
|
|
state_hub_task_id: "efb153cd-3963-5d40-b327-1b1cee793cb6"
|
|
```
|