Skip to content

[docs] Update documentation for features from 2026-10-20 #70500

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
github-actions merged 1 commit into from
Oct 20, 2026
Merged

[docs] Update documentation for features from 2026-10-20 #70500

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 2 additions & 0 deletions docs/src/content/docs/reference/engines.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -571,6 +571,8 @@ Threat detection runs default `max-retries` to **0** instead of inheriting the h

The harness logs the configured and resolved model and records them as a `model.resolution` event in `usage/aw_session.jsonl`, so audits show when an alias moved to a different model. Gemini and declarative engines do not resolve aliases yet: they receive the configured model unchanged, and the compiler warns when one of them is given an alias. Built-in aliases are listed in [Model Aliases](/gh-aw/reference/model-tables/).

If Pi rejects a sub-agent's model during its startup availability check, the resolution record for every sub-agent using that model reports `status: failed`, `cause: model_unavailable`, and `resolved_model: null`. This replaces the earlier successful resolution record in both routed and unrouted runs; startup stops before the sub-agent executes.

### Ordered model fallback (`fallback-models`)

All built-in engines accept an ordered list of alternative models. With AWF v0.28.31 or newer and a compatible API-proxy image, concrete same-provider chains compile to `apiProxy.fallbackModels`. AWF retries the failed inference request with the next model without restarting the agent or replaying its task, and applies the same model-policy and budget guards to each alternative. An explicit chain disables AWF's automatic middle-power model substitution so only configured alternatives are tried.
Expand Down
17 changes: 14 additions & 3 deletions docs/src/content/docs/setup/cli.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -473,7 +473,7 @@ gh aw lint --rules secret-leak --debug --json # Include incomplete-analysi

By default, shellcheck and pyflakes integrations are disabled to reduce noise for generated `run:` scripts. Built-in actionlint ignore patterns cover gh-aw-specific extensions such as `job.workflow_*` context properties and the `copilot-requests` permission scope.

`--rules` accepts comma-separated rule IDs; `--rules=` selects all registered rules without actionlint. Unknown IDs fail explicitly. `--json` requires `--rules`; `--shellcheck` and `--pyflakes` apply only to actionlint. In the default mode, both linters run even if one reports errors, and either linter's failure produces a nonzero exit status.
`--rules` accepts comma-separated rule IDs; `--rules=` selects all registered rules without actionlint. Unknown IDs fail explicitly. `--json` requires `--rules` and has no `-j` shorthand; `--shellcheck` and `--pyflakes` apply only to actionlint. In the default mode, both linters run even if one reports errors, and either linter's failure produces a nonzero exit status.

Secret-leak coverage warnings are hidden by default in text and JSON output, unless JavaScript analysis is explicitly enabled. Use `--debug` to include them otherwise. Error findings, analysis failures, and other linters' warnings remain visible. This is a presentation setting: the AST runner retains all coverage diagnostics, and an empty CLI result is not proof of complete analysis.

Expand Down Expand Up @@ -1254,9 +1254,20 @@ gh aw work-queue explain --repo owner/repo --work-id my-work
gh aw work-queue stats --repo owner/repo
```

**Persistent options:** `--branch` (queue branch, default `work-queue`), `--json/-j`, `--repo/-r` (`owner/repo`), `--request-id` (publication handle or committed request selector), `--storage` (`git` only).
**Persistent options:** `--branch` (queue branch, default `work-queue`), `--json/-j`, `--repo/-r` (explicit `owner/repo` for repository operations; no host prefix or current-repository default), `--request-id` (publication handle or committed request selector), `--storage` (`git` only).

**Subcommands:** `cancel-claim`, `cancel-work`, `compact`, `control`, `deploy`, `dispatch-next`, `evidence`, `explain`, `inspect`, `policy`, `reconcile`, `replay`, `reprioritize`, `state`, `stats`, `submit-work`, `submit-graph`, `trace`, `tui`. Use `gh aw work-queue <subcommand> --help` for subcommand-specific options.
**Subcommands:** `cancel-claim`, `cancel-work`, `clear`, `compact`, `control`, `deploy`, `dispatch`, `dispatch-next`, `evidence`, `explain`, `inspect`, `policy`, `reconcile`, `replay`, `reprioritize`, `state`, `stats`, `submit-work`, `submit-graph`, `trace`, `tui`. Use `gh aw work-queue <subcommand> --help` for all subcommand-specific options.

| Subcommand | Options and behavior |
| --- | --- |
| `submit-work` | `--graph-id` sets a nonempty graph namespace (default: canonical payload SHA-256); `--node-key` sets a nonempty node identity within the graph (default: `root`). `--fairness-key` selects an entitled accounting key (default: one shared bucket). |
| `dispatch-next` | `--max-bytes` limits assignment bytes (default: 49,152), subject to policy and host limits. Grants and reserves assignments without launching workflows. |
| `dispatch <dispatcher>` | Starts a verified dispatcher at the repository's default-branch snapshot revision. `--dry-run` previews the route without granting Claims or launching workflows; `--wait` waits for the dispatcher, not verification of worker Results. |
| `control` | `--name credential_generation` requires `--generation`, an opaque identifier for an equivalent-scope credential generation, never a secret. Pause controls (`admission_paused`, `grants_paused`) require explicit `--paused=true` or `--paused=false`. All controls require `--reason`. |
| `reconcile` | `--cancel-reserved` requires `--dispatch-id` and cancels only when compare-and-set proves no launch-start marker exists. It cannot force-release a possible native run and cannot be combined with `--work-id`. |
| `clear` | Requires maintain or admin permission and `--reason CODE`. `--dry-run` validates and previews without publishing; `--hard` resets a paused, fully settled ledger. All modes, including previews, refuse CI. Mutations require interactive confirmation. |

Normal `clear` cancels unfinished snapshot Work without pending deliveries and preserves completed Work, history, fairness debt, and native reservations. It neither stops native runs nor pauses admission or grants. `--hard` requires admission and grants already paused, all reservations released, and all effects settled; it forgets active identities and fairness debt while retaining Git ancestry, so old identities can be submitted again and repeat effects. See [clearing unfinished Work](/gh-aw/reference/work-queue/#clear-unfinished-work) for limits and operator procedures.

#### `version`

Expand Down
Loading