docks-kit 0.22.20 → 0.22.23
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/AGENTS.md +5 -5
- package/cli/docs/command-guards.md +5 -4
- package/cli/docs/omp-context.md +13 -9
- package/cli/docs/omp-models.md +90 -69
- package/cli/src/commands/omp.ts +25 -73
- package/cli/src/engine-native/harnesses.ts +5 -11
- package/cli/src/engine-native/kitDb.ts +7 -7
- package/cli/src/engine-native/ompOverlay.ts +23 -69
- package/cli/src/engine-native/ompRemovals.ts +47 -8
- package/cli/src/engine-native/sharedTypes.d.ts +0 -2
- package/cli/src/generated/sotPayload.ts +5 -5
- package/package.json +1 -1
package/AGENTS.md
CHANGED
|
@@ -87,16 +87,16 @@ omp SoT notes:
|
|
|
87
87
|
- `SoT/.omp/AGENTS.md`, `config.yml`, `models.yml`, and `mcp.json` deploy to `~/.omp/agent/`.
|
|
88
88
|
- `SoT/.omp/intercom.json` deploys to `$PI_CODING_AGENT_DIR/intercom/config.json`. The default root is `~/.pi/agent`. Under a named omp profile (`OMP_PROFILE` or `PI_PROFILE`), it deploys to `<profile agentDir>/intercom/config.json`, because omp exports `PI_CODING_AGENT_DIR` as the profile agent directory.
|
|
89
89
|
- `ompFileDeploy.ts syncMergedYaml` deep-merges `config.yml` through `ompYaml.ts mergeOmpConfig` and `models.yml` through `mergeOmpModels`. Both wrap one generic mapping merge; only the config wrapper prunes stale `retry.fallbackChains` wildcards and merges `bash.patterns`.
|
|
90
|
-
- `SoT/.omp/config.yml` `bash.patterns` holds the destructive-command guards. omp's `CRITICAL_BASH_PATTERNS` check returns a bare safety override before a matching prompt rule, and `tools.approvalMode: yolo` ignores a bare override, so a prompt on a critical command becomes an allow, in subagents too (`omp://approval-mode.md`). Every `rm` shape that check can flag is therefore a deny: recursive `rm`, `sudo` anywhere before `rm` (also by path), any `rm` with an `r` or `f` option before an absolute path, and `--no-preserve-root`. By owner decision (0.20.9), the recursive-`rm` deny also covers any prefix (a `* rm ` copy of each bare cluster rule, so `env X=1 rm -rf b`, `xargs rm -rf`, `find . -exec rm -rf {} +`, and `git rm -r dir` are denied) and the `-ri`, `-ir`, `-Ri`, and `-iR` clusters for bare `rm`, `*/rm`, and `* rm`. By owner decision (0.21.0), 17 more denies cover bulk removal that can empty the home directory: `find -delete` (bare, `*/find`, `* find`), `-exec`, `-execdir`, `-ok`, and `-okdir` with `rm` or `*/rm`, and `xargs rm` in six shapes. No glob can express "inside the temporary directory", so `docks-kit scratch remove` checks that path in code. A `$` or a backtick anywhere after `git push` in the same command is a deny, not a prompt (0.22.5, #120). A variable can expand to a `+` refspec or `--force`, and the matcher reads the text before the shell expands it. The agent reads the refusal and writes the ref literally, so no person is asked. The two rules are `*git push *$*` and ``*git push *`*``. Their leading `*` keeps a shell keyword such as `do`, a wrapper such as `env`, or a path before `git` from hiding the push. By owner decision (0.22.8, #126), the kit ships no omp prompt rule: `bash.patterns` holds only the denies plus the `trash *` allow, so every command that no deny matches runs without a question, in subagents too. Since 0.20.9 this run-without-question behavior covers a non-denied `rm`, `git rm`, `git worktree remove`, and `gio trash`. Since 0.21.0 it also covers `docks-kit scratch remove`. Since 0.22.8 it also covers `git clean`, `git reset --hard`, a hook skip, and every `git push` that no deny matches. The force-push denies match only `git push --force` or `-f` directly before `origin main` or `origin master`, so another spelling, such as `git push origin main --force`, runs. A prompt rule that the user adds survives sync and still asks. `SoT/.omp/AGENTS.md` carries the agent rules for these commands, and it tells the agent to write each `git push` ref literally and to push the current branch with `git push -u origin HEAD`. Claude Code `permissions.ask` and Codex `docks.rules` keep their ask tiers. The list order is deny, then `trash *`. omp takes the first match. `ompYaml.ts mergeBashPatterns, grouped rule merge` keeps user rules and groups all rules as kit deny, user deny, user prompt, kit prompt, other user rules, kit allow, so a user deny or prompt precedes every kit allow. It treats a deployed rule equal to a kit rule (approval case-folded, `match` whitespace collapsed) as the kit copy, so a re-sync is a no-op. A rule the kit stops shipping stays as a user rule, so retire one through `ompRemovals.ts OMP_RETIRED_BASH_RULES`. `cli/test/lib/ompBashPatterns.ts` ports omp 18.17.0's matcher, critical patterns, and yolo resolution; `cli/test/unit/ompBashPatterns.test.ts` asserts each command's final outcome and the merge. omp subagents reject every `prompt`, and the `eval` tool stays unrestricted by owner decision.
|
|
90
|
+
- `SoT/.omp/config.yml` `bash.patterns` holds the destructive-command guards. omp's `CRITICAL_BASH_PATTERNS` check returns a bare safety override before a matching prompt rule, and omp's default `tools.approvalMode: yolo` ignores a bare override, so a prompt on a critical command becomes an allow, in subagents too (`omp://approval-mode.md`). Every `rm` shape that check can flag is therefore a deny: recursive `rm`, `sudo` anywhere before `rm` (also by path), any `rm` with an `r` or `f` option before an absolute path, and `--no-preserve-root`. By owner decision (0.20.9), the recursive-`rm` deny also covers any prefix (a `* rm ` copy of each bare cluster rule, so `env X=1 rm -rf b`, `xargs rm -rf`, `find . -exec rm -rf {} +`, and `git rm -r dir` are denied) and the `-ri`, `-ir`, `-Ri`, and `-iR` clusters for bare `rm`, `*/rm`, and `* rm`. By owner decision (0.21.0), 17 more denies cover bulk removal that can empty the home directory: `find -delete` (bare, `*/find`, `* find`), `-exec`, `-execdir`, `-ok`, and `-okdir` with `rm` or `*/rm`, and `xargs rm` in six shapes. No glob can express "inside the temporary directory", so `docks-kit scratch remove` checks that path in code. A `$` or a backtick anywhere after `git push` in the same command is a deny, not a prompt (0.22.5, #120). A variable can expand to a `+` refspec or `--force`, and the matcher reads the text before the shell expands it. The agent reads the refusal and writes the ref literally, so no person is asked. The two rules are `*git push *$*` and ``*git push *`*``. Their leading `*` keeps a shell keyword such as `do`, a wrapper such as `env`, or a path before `git` from hiding the push. By owner decision (0.22.8, #126), the kit ships no omp prompt rule: `bash.patterns` holds only the denies plus the `trash *` allow, so every command that no deny matches runs without a question, in subagents too. Since 0.20.9 this run-without-question behavior covers a non-denied `rm`, `git rm`, `git worktree remove`, and `gio trash`. Since 0.21.0 it also covers `docks-kit scratch remove`. Since 0.22.8 it also covers `git clean`, `git reset --hard`, a hook skip, and every `git push` that no deny matches. The force-push denies match only `git push --force` or `-f` directly before `origin main` or `origin master`, so another spelling, such as `git push origin main --force`, runs. A prompt rule that the user adds survives sync and still asks. `SoT/.omp/AGENTS.md` carries the agent rules for these commands, and it tells the agent to write each `git push` ref literally and to push the current branch with `git push -u origin HEAD`. Claude Code `permissions.ask` and Codex `docks.rules` keep their ask tiers. The list order is deny, then `trash *`. omp takes the first match. `ompYaml.ts mergeBashPatterns, grouped rule merge` keeps user rules and groups all rules as kit deny, user deny, user prompt, kit prompt, other user rules, kit allow, so a user deny or prompt precedes every kit allow. It treats a deployed rule equal to a kit rule (approval case-folded, `match` whitespace collapsed) as the kit copy, so a re-sync is a no-op. A rule the kit stops shipping stays as a user rule, so retire one through `ompRemovals.ts OMP_RETIRED_BASH_RULES`. `cli/test/lib/ompBashPatterns.ts` ports omp 18.17.0's matcher, critical patterns, and yolo resolution; `cli/test/unit/ompBashPatterns.test.ts` asserts each command's final outcome and the merge. omp subagents reject every `prompt`, and the `eval` tool stays unrestricted by owner decision.
|
|
91
91
|
- `ompSync.ts` runs `ompRemovals.ts syncOmpRemovals, retired-key inventory` right after the config merge, and that pass force-prunes retired kit-owned keys from `~/.omp/agent/config.yml` on every sync, without `--reconcile`. The pass is required because `mergeOmpConfig` is additive, so removing a key from the SoT alone never removes it from a deployed file. A key retired with a recorded value is pruned only while the deployed value still matches that value, so a user edit survives; a key retired outright, such as `providers.webSearchOrder`, is pruned at any value. The same pass removes each `bash.patterns` rule listed in `OMP_RETIRED_BASH_RULES`: the 39 file-removal rules retired in 0.20.9, the four `git push` prompts for a `$` or backtick word that 0.22.5 replaced with denies, and the 492 prompt rules retired in 0.22.8 (`RETIRED_PROMPT_MATCHES`). It compares with the merge's kit-rule normalization (approval case-folded, `match` whitespace collapsed), so a user rule with a different `match` or `approval` survives.
|
|
92
|
-
- `cycleOrder` ends with `astra` as its fifth stop. `modelRoles.astra` is `openai-codex/gpt-6-astra:xhigh`, and `modelTags.astra` is visible. Astra and Fable fall back to each other through concrete selectors. `modelRoles.fable`
|
|
93
|
-
- `modelRoles.task` is `anthropic/claude-opus-5-5:high`, and `retry.fallbackChains.task` falls back to `openai-codex/gpt-6.1-sol:high`. `task.enableEffort
|
|
94
|
-
- The
|
|
92
|
+
- `cycleOrder` ends with `astra` as its fifth stop. `modelRoles.astra` is `openai-codex/gpt-6-astra:xhigh`, and `modelTags.astra` is visible. Astra and Fable fall back to each other through concrete selectors. `modelRoles.fable` is `anthropic/claude-fable-5-2:medium`, with visible `modelTags.fable`. The hidden `switch_fable` role uses the same Fable selector and keeps an empty fallback chain.
|
|
93
|
+
- `modelRoles.task` is `anthropic/claude-opus-5-5:high`, and `retry.fallbackChains.task` falls back to `openai-codex/gpt-6.1-sol:high`. The kit relies on omp's default `task.enableEffort: false`: a caller `effort` outranks the role's level suffix, and `hi` maps to the top of the model ladder, so with the field on, `hi` spawns ran at `task.maxEffort` (`xhigh`) instead of the role level. With the field off, every spawn runs its role level, and `task.maxEffort` bounds only a per-call `model` with a `:level` suffix. By owner decision (0.22.22), the omp advisor is removed: the SoT has no `advisor` block, `advisor` role, or `advisor` fallback chain. The plan-lifecycle verification loop replaces a second model on every turn. A session that runs `/advisor on` takes its advisor model from the `slow` role. `smol`, `commit`, and `tiny` use `openai-codex/gpt-6-luna`. The kit-defined hidden `reviewer` role is `openai-codex/gpt-6.1-terra:high`. Its retry fallback is `anthropic/claude-opus-5-5:high`. All four reviewer entries in `task.agentModelOverrides` point to `@reviewer`. `task.agentServiceTierOverrides` sets the same four agents to `inherit`, so they take the live tier of the main session for each provider. The user set the effort `high` and chose to follow the fast toggle (#134); the review benchmark of #130 chose the model. `ompOverlay.ts ROLE_ORDER, FALLBACK_ORDER` list `reviewer`, so a free session also replaces it. omp bundles the `reviewer` and `security-reviewer` agents, and the `plan-lifecycle` plugin ships `code-reviewer` and `plan-reviewer`. The `SoT/toolchain.json` omp floor is 18.4.4, the oldest release the kit tested with `gpt-6.1-sol`. That release lists and serves the model through the `openai-codex` provider. `cli/docs/omp-models.md` records the checks.
|
|
94
|
+
- The six Opus 5.5 roles (`default`, `slow`, `plan`, `designer`, `vision`, `task`) and the `smol`, `tiny`, `commit`, and `reviewer` retry chains use `anthropic/claude-opus-5-5` at the levels the role map records. `cli/docs/omp-models.md` carries the Artificial Analysis capture behind the role map, read 2026-09-29 at Intelligence Index v4.3.2 and Coding Agent Index v1.5 from the AA comparison-page metric tables. The `reviewer`, `fable`, and `switch_fable` rows are the exceptions: they come from v5.0.1 captures read 2026-11-05 (GPT-6.1 Terra) and 2026-11-12 (Fable 5.2), and they compare only with other v5.0.1 figures. Opus 5.5 max has the highest v4.3.2 index in that topic, at 58. AA has measured speed and latency for Opus 5.5 max and every GPT-6.1 Sol level; GPT-6 Luna medium is the only unmeasured speed and latency row.
|
|
95
95
|
- `modelRoles.web` is `web/firecrawl` and `retry.fallbackChains.web` carries the explicit 20-entry provider order. The kit declares both keys because the legacy `providers.webSearchOrder` key is retired: omp expands it in memory into these two keys and then drops it, and never writes that expansion back to disk. An explicit chain replaces omp's built-in web order wholesale, so every entry left out is a provider omp never tries. The owner removed the seven entries that named older models (Gemini 2.5 Flash, Claude Haiku 4.5, GPT-5.6, GPT-5.5, and Grok 4.5); keep every other provider.
|
|
96
96
|
- `SoT/.omp/models.yml` declares Astra's full `low, medium, high, xhigh, max` ladder with `defaultLevel: xhigh` as the worked provider ladder-override example. The temporary `anthropic.modelOverrides.claude-opus-5-5` block is gone, because the shared catalog now publishes the same limits, ladder, and prices. The removal switches omp from the block's `thinking.mode: effort` (a `budget_tokens` value per level from `ANTHROPIC_THINKING`, no effort value) to the catalog's `anthropic-adaptive` (adaptive thinking with the role's effort level); every kit Opus selector names its level, so the block's bare-selector `defaultLevel: high` is not needed. `ompRemovals.ts syncOmpModelRemovals, retired-block inventory` prunes it from `~/.omp/agent/models.yml` after the models merge, but only while the deployed block still equals the shipped one. `ompYaml.ts mergeOmpModels` preserves deployed-only keys in `~/.omp/agent/models.yml`, because a user file may carry provider credentials. Whole-file replacement is wrong.
|
|
97
97
|
- `SoT/.omp/AGENTS.md` carries the rule `Please remove all mannered prose.` Anthropic's Fable 5.1 prompting guide documents mannered prose as a Fable 5.1 behavior and gives that sentence as its short-version fix: https://platform.claude.com/docs/en/build-with-claude/prompt-engineering/prompting-claude-fable-5-1
|
|
98
98
|
- `cli/docs/omp-models.md` (topic `omp-models`) records the role map rationale and the Artificial Analysis snapshot behind it. Model choices change with published benchmarks, so update that topic in the same commit as a role change.
|
|
99
|
-
- `SoT/.omp/config.yml`
|
|
99
|
+
- `SoT/.omp/config.yml` carries only keys whose value differs from the omp default. When a kit value equals the omp default, remove the key from the SoT. Add it with its shipped value to `ompRemovals.ts OMP_RETIRED_VALUES`, because `ompYaml.ts mergeOmpConfig` is additive and only that pass prunes the deployed copy. For example, the kit ships no `compaction.thresholdTokens`, and sync prunes a deployed `-1` and the former `231200` pin. `cli/docs/omp-context.md` (topic `omp-context`) carries the reserve-based derivation and the measured evidence, so update that topic in the same commit as any compaction-setting change.
|
|
100
100
|
- Sync materializes the kit omp plugin `docks-kit-omp` from `SoT/.omp/plugin/` to `~/.docks-kit/omp-plugin` (`ompKitPlugin.ts syncKitPlugin`). It sets directories to 0700, files to 0600, and the `package.json` `version` to the kit version, and it removes stale files. It keeps no backups, because the directory is kit-owned. `ompPlugins.ts syncPlugins, kit plugin link` then runs `omp plugin link <dir>`. It skips the link when the `omp plugin list --json` npm row `docks-kit-omp` already has a `path` that resolves to that directory. `--skip-plugin-refresh` does not skip the link. The `hygiene.ts` extension registers `scratch_new`, `scratch_remove`, and `tidy_report`. It adds a reminder after each bash command that created a resource, and it checks at session stop that those resources are gone. At a main-session start, it claims `~/.cache/docks-kit/tidy.stamp.claim` and runs `docks-kit tidy --apply` in the background at most once per 12 hours. Every extension subprocess runs without a shell. The `pluginUpgrades.ts` extension reads the omp plugin upgrade log. In a tool result of each main session, it adds one reminder for the logged upgrades that the session did not get before. It names only the upgrades that finished after its omp process started. The reminder tells the agent to ask the user to run `/reload-plugins`. It then tells the agent to read each plugin skill again through `skill://` before it runs a skill script. It asks for a restart instead when an upgraded plugin ships hooks, extension modules, or non-MCP tools. omp imports extensions only at startup. An omp process that started before a sync deployed `pluginUpgrades.ts` did not load it, so its sessions get no reminder. The plugin also ships the `workspace-hygiene` skill and the `/cleanup` command. The plugin, the deployed rules, and the sweep call `docks-kit scratch` and `docks-kit tidy`. So a checkout sync must not run before the global `docks-kit` carries those commands.
|
|
101
101
|
- `task.agentIdleTtlMs` is `120000`, so omp parks an idle subagent after 2 minutes instead of 7. `browser.idleCloseSec` is `600`.
|
|
102
102
|
- By owner decision (0.21.0), resources on the workstation are always development, so `docks-kit tidy` sweeps them without a question. The omp agent prompt, the `workspace-hygiene` skill, `/cleanup`, and the extension reminders limit the cleanup of a remote resource (reached through SSH or a provider CLI such as `wrangler`, `vercel`, `kubectl`, or a remote Docker host) to targets named local, development, dev, test, homolog, homologation, staging, or preview, the plan-lifecycle non-production definition. A production cleanup needs an `ask` immediately before it, must be rare, and a `yolo` invocation does not answer that ask. Prompt rules cannot force an agent to ask, so this rule has no `bash.patterns` guard.
|
|
@@ -16,8 +16,8 @@ ask tier is accepted.
|
|
|
16
16
|
|
|
17
17
|
omp is the exception: by owner decision (0.22.8, #126), the kit ships no omp
|
|
18
18
|
prompt rule. The kit's `bash.patterns` holds only deny rules and the `trash *`
|
|
19
|
-
allow. Under
|
|
20
|
-
that no deny matches without a question, in the main session and in
|
|
19
|
+
allow. Under omp's default `tools.approvalMode: yolo`, omp runs every
|
|
20
|
+
command that no deny matches without a question, in the main session and in
|
|
21
21
|
subagents. This includes any `rm` that omp does not deny, `git clean`,
|
|
22
22
|
`git reset --hard`, a force push that no deny matches, and a git command
|
|
23
23
|
that skips the repository's hooks. A prompt rule that the user adds to
|
|
@@ -175,8 +175,9 @@ omp has a built-in check for critical commands. For example, it flags:
|
|
|
175
175
|
- `sudo rm`, also when `sudo` is called by its path.
|
|
176
176
|
|
|
177
177
|
For a critical command, omp returns a bare safety override before it returns
|
|
178
|
-
a matching prompt rule. Under
|
|
179
|
-
ignores a bare override, so the command runs. This applies in subagents
|
|
178
|
+
a matching prompt rule. Under omp's default `tools.approvalMode: yolo`,
|
|
179
|
+
omp ignores a bare override, so the command runs. This applies in subagents
|
|
180
|
+
too.
|
|
180
181
|
Only a deny rule stops a critical command (`omp://approval-mode.md`).
|
|
181
182
|
|
|
182
183
|
The kit therefore denies every `rm` shape that the check can flag:
|
package/cli/docs/omp-context.md
CHANGED
|
@@ -1,7 +1,7 @@
|
|
|
1
1
|
# omp context and compaction
|
|
2
2
|
|
|
3
|
-
This topic records why `SoT/.omp/config.yml`
|
|
4
|
-
|
|
3
|
+
This topic records why `SoT/.omp/config.yml` ships no compaction threshold,
|
|
4
|
+
and what omp's reserve-based default computes on each model lane.
|
|
5
5
|
|
|
6
6
|
## How omp resolves the trigger
|
|
7
7
|
|
|
@@ -62,17 +62,20 @@ No session ran past its window, and omp raised no warning on either lane.
|
|
|
62
62
|
Under the reserve-based default none of the 75 Anthropic compactions would
|
|
63
63
|
have fired, because no session reached 850,000.
|
|
64
64
|
|
|
65
|
-
##
|
|
65
|
+
## How sync removes the deployed pin
|
|
66
|
+
|
|
67
|
+
The kit no longer ships `compaction.thresholdTokens`. omp's schema default
|
|
68
|
+
`-1` applies, and that sentinel selects the reserve-based branch.
|
|
66
69
|
|
|
67
70
|
`ompYaml.ts` `mergeOmpConfig` is additive. Its `mergeMappings, deployed-key
|
|
68
71
|
retention loop` returns every deployed key that the SoT omits to the merged
|
|
69
72
|
output, and the only drop predicate is `dropFallbackWildcard`, which is scoped
|
|
70
|
-
to `retry.fallbackChains`.
|
|
73
|
+
to `retry.fallbackChains`. A key deleted from the SoT therefore stays in every
|
|
74
|
+
already-deployed `~/.omp/agent/config.yml`.
|
|
71
75
|
|
|
72
|
-
|
|
73
|
-
|
|
74
|
-
|
|
75
|
-
next sync.
|
|
76
|
+
`ompRemovals.ts OMP_RETIRED_VALUES` therefore records both values that the kit
|
|
77
|
+
shipped: the former `231200` pin and the later `-1`. Sync prunes the deployed
|
|
78
|
+
key only while it holds one of these values, so a user value survives.
|
|
76
79
|
|
|
77
80
|
Verify the deployed value after a sync:
|
|
78
81
|
|
|
@@ -80,7 +83,8 @@ Verify the deployed value after a sync:
|
|
|
80
83
|
omp config get compaction.thresholdTokens
|
|
81
84
|
```
|
|
82
85
|
|
|
83
|
-
The command must print `-1
|
|
86
|
+
The command must print `-1`, the omp default, unless the user set another
|
|
87
|
+
value.
|
|
84
88
|
|
|
85
89
|
## The clamp hazard a fixed pin carries
|
|
86
90
|
|
package/cli/docs/omp-models.md
CHANGED
|
@@ -12,13 +12,12 @@ against.
|
|
|
12
12
|
| `slow` | `anthropic/claude-opus-5-5` | xhigh | 56 | $3.46 | 136.30 s |
|
|
13
13
|
| `plan` | `anthropic/claude-opus-5-5` | xhigh | 56 | $3.46 | 136.30 s |
|
|
14
14
|
| `task` | `anthropic/claude-opus-5-5` | high | 54 | $1.82 | 52.85 s |
|
|
15
|
-
| `advisor` | `anthropic/claude-opus-5-5` | medium | 51 | $1.34 | 21.87 s |
|
|
16
15
|
| `designer` | `anthropic/claude-opus-5-5` | high | 54 | $1.82 | 52.85 s |
|
|
17
16
|
| `vision` | `anthropic/claude-opus-5-5` | medium | 51 | $1.34 | 21.87 s |
|
|
18
17
|
| `smol` / `commit` | `openai-codex/gpt-6-luna` | medium | 29 | $0.02 | n/a |
|
|
19
18
|
| `tiny` | `openai-codex/gpt-6-luna` | low | 21 | $0.0045 | 2.18 s |
|
|
20
|
-
| `fable` | `anthropic/claude-fable-5-
|
|
21
|
-
| `switch_fable` | `anthropic/claude-fable-5-
|
|
19
|
+
| `fable` | `anthropic/claude-fable-5-2` | medium | 56 (v5.0.1) | $3.80 (v5.0.1) | 11.23 s (v5.0.1) |
|
|
20
|
+
| `switch_fable` | `anthropic/claude-fable-5-2` | medium | 56 (v5.0.1) | $3.80 (v5.0.1) | 11.23 s (v5.0.1) |
|
|
22
21
|
| `astra` | `openai-codex/gpt-6-astra` | xhigh | 52 | $2.31 | 126.90 s |
|
|
23
22
|
| `reviewer` | `openai-codex/gpt-6.1-terra` | high | 42 (v5.0.1) | $0.26 (v5.0.1) | 10.91 s (v5.0.1) |
|
|
24
23
|
| `web` | `web/firecrawl` | n/a | n/a | n/a | n/a |
|
|
@@ -27,10 +26,11 @@ The table reports the measured Artificial Analysis figures for each assigned
|
|
|
27
26
|
model and level. It states no motive that the config or omp's own
|
|
28
27
|
documentation does not establish. AA has not measured output speed or latency
|
|
29
28
|
for GPT-6 Luna medium, so that row carries `n/a` for TTFT. AA measures no web
|
|
30
|
-
search provider, so the `web` row carries no figures. The `reviewer`
|
|
31
|
-
from
|
|
32
|
-
predates GPT-6.1 Terra
|
|
33
|
-
GPT-6.1 Terra
|
|
29
|
+
search provider, so the `web` row carries no figures. The `reviewer`, `fable`,
|
|
30
|
+
and `switch_fable` rows come from later captures at Intelligence Index v5.0.1.
|
|
31
|
+
The 2026-09-29 capture predates GPT-6.1 Terra and Fable 5.2. Compare those
|
|
32
|
+
rows only with other v5.0.1 figures (see GPT-6.1 Terra and Claude Fable 5.2
|
|
33
|
+
below). A review benchmark chose the model of the `reviewer` role
|
|
34
34
|
(DocksDocks/public#130), not these figures. The tier of that role follows the
|
|
35
35
|
main session (DocksDocks/public#134).
|
|
36
36
|
|
|
@@ -116,7 +116,7 @@ missed two pre-registered bounds: escaped difference up to 3.26 (bound 2) and
|
|
|
116
116
|
recall difference down to -0.07 (bound -0.05). Both intervals include 0, and
|
|
117
117
|
the user judged the difference as noise.
|
|
118
118
|
|
|
119
|
-
`retry.fallbackChains.astra` holds `anthropic/claude-fable-5-
|
|
119
|
+
`retry.fallbackChains.astra` holds `anthropic/claude-fable-5-2:medium`.
|
|
120
120
|
`retry.fallbackChains.fable` holds `openai-codex/gpt-6-astra:xhigh`.
|
|
121
121
|
Each deliberate cycle stop falls to the other vendor. Without these explicit
|
|
122
122
|
chains, `retry.fallbackChains.default` would send either stop to
|
|
@@ -124,22 +124,21 @@ chains, `retry.fallbackChains.default` would send either stop to
|
|
|
124
124
|
Chain entries are concrete selectors, not role aliases, so this pair cannot
|
|
125
125
|
recurse. The hidden `switch_fable` chain stays empty.
|
|
126
126
|
|
|
127
|
-
###
|
|
128
|
-
|
|
129
|
-
|
|
130
|
-
|
|
131
|
-
|
|
132
|
-
|
|
133
|
-
|
|
134
|
-
|
|
135
|
-
|
|
136
|
-
|
|
137
|
-
|
|
138
|
-
|
|
139
|
-
|
|
140
|
-
|
|
141
|
-
|
|
142
|
-
`retry.fallbackChains.advisor`.
|
|
127
|
+
### Advisor removed
|
|
128
|
+
|
|
129
|
+
By owner decision (2026-11-12, 0.22.22), the kit sets no omp advisor.
|
|
130
|
+
`SoT/.omp/config.yml` has no `advisor` block, no `modelRoles.advisor`, and no
|
|
131
|
+
`retry.fallbackChains.advisor`. omp keeps the advisor off by default. The
|
|
132
|
+
plan-lifecycle verification loop takes its place: acceptance checks, a fresh
|
|
133
|
+
code review of the diff, and a round limit. Since 2026-11-01, 33 main
|
|
134
|
+
sessions had an advisor log, and no subagent session had one. A session that
|
|
135
|
+
runs `/advisor on` takes its advisor model from the `slow` role, because no
|
|
136
|
+
`advisor` role is set.
|
|
137
|
+
|
|
138
|
+
History: on 2026-09-24, with omp 18.3.0, an advisor on
|
|
139
|
+
`openai-codex/gpt-6-sol:medium` looped and made 1,313 requests for 140
|
|
140
|
+
main-agent requests. The kit then ran the advisor on Opus 5.5 `medium` and
|
|
141
|
+
excluded every GPT model from that role until this removal.
|
|
143
142
|
|
|
144
143
|
`modelRoles.web` is `web/firecrawl`, and `retry.fallbackChains.web` lists the
|
|
145
144
|
explicit 20-entry provider order that follows it. The two keys replace the
|
|
@@ -254,6 +253,34 @@ selector, so no kit role depends on that default. Runs on
|
|
|
254
253
|
`anthropic/claude-opus-5-5:high` and `:max` after the prune answered with
|
|
255
254
|
exit 0.
|
|
256
255
|
|
|
256
|
+
### Claude Fable 5.2 (Anthropic) - `anthropic/claude-fable-5-2`
|
|
257
|
+
|
|
258
|
+
Source: `https://artificialanalysis.ai`, read on 2026-11-12 at Intelligence
|
|
259
|
+
Index v5.0.1, from the metric table of
|
|
260
|
+
`/models/comparisons/<level-slug>-vs-gpt-5-6-sol-high`. AA serves `max` under
|
|
261
|
+
the bare model slug. These figures use another index version than the rest of
|
|
262
|
+
this snapshot, so they do not compare with the v4.3.2 rows above. Compare them
|
|
263
|
+
with GPT-6.1 Terra below.
|
|
264
|
+
|
|
265
|
+
| Level | Index | Cost/task | Tokens/task | Index tokens | Speed t/s | TTFT s | TB 4.0 |
|
|
266
|
+
|---|---:|---:|---:|---:|---:|---:|---:|
|
|
267
|
+
| max | 65 | $26.54 | 274k | 275M | 98 | 262.29 | 60% |
|
|
268
|
+
| xhigh | 62 | $14.83 | 162k | 144M | 102 | 85.87 | 62% |
|
|
269
|
+
| high | 62 | $9.08 | 98k | 85M | 97 | 27.98 | 60% |
|
|
270
|
+
| medium | 56 | $3.80 | 43k | 41M | 93 | 11.23 | 55% |
|
|
271
|
+
| low | 54 | $3.10 | 36k | 35M | 97 | 8.42 | 55% |
|
|
272
|
+
|
|
273
|
+
Price: $8.00 in, $40.00 out, $0.20 cache hit per 1M. The omp catalog lists
|
|
274
|
+
the same prices, a $10.00 cache write, a 1M context, and 128K maximum output.
|
|
275
|
+
All AA levels run with default fallback. On 2026-11-12, omp 18.19.0 listed
|
|
276
|
+
`claude-fable-5-2` at low, medium, high, xhigh, and max, and an
|
|
277
|
+
`omp -p --mode json --model anthropic/claude-fable-5-2:medium` run recorded
|
|
278
|
+
`claude-fable-5-2` as the serving model.
|
|
279
|
+
|
|
280
|
+
`fable` and `switch_fable` moved from Fable 5.1 to Fable 5.2 at the same
|
|
281
|
+
level, `medium`. The Fable 5.1 table below stays as the previous-generation
|
|
282
|
+
baseline at v4.3.2.
|
|
283
|
+
|
|
257
284
|
### Claude Fable 5.1 (Anthropic) - `anthropic/claude-fable-5-1`
|
|
258
285
|
|
|
259
286
|
| Level | Index | Cost/task | Tokens/task | Index tokens | Speed t/s | TTFT s | TB 4.0 |
|
|
@@ -442,18 +469,18 @@ quick answer.
|
|
|
442
469
|
`thinking-level: auto`. It resolves Opus 5.5 high. The `:high` suffix on
|
|
443
470
|
`modelRoles.task` is an explicit level, so it outranks the agent's `auto`,
|
|
444
471
|
and the auto classifier does not run for a `task` spawn.
|
|
445
|
-
- `task.enableEffort`
|
|
446
|
-
every spawn runs the level of its agent's role. When the
|
|
447
|
-
caller can pass `effort: lo`, `med`, or `hi`, and caller
|
|
448
|
-
role suffix. omp maps it onto the model's own ladder,
|
|
449
|
-
`lo` selects the lowest level, `med` the lower middle,
|
|
450
|
-
(`resolveTaskEffortLevel` in omp's
|
|
451
|
-
omp 18.8.5). Then omp caps the result at
|
|
452
|
-
`low, medium, high, xhigh, max` ladder, `hi` gave
|
|
453
|
-
`xhigh` cap, so many `task` spawns ran Opus 5.5 xhigh
|
|
454
|
-
role suffix. On 2026-11-05, a fresh `omp -p` run on omp
|
|
455
|
-
field off had no `effort` field, and its `task` child
|
|
456
|
-
`resolvedModel: anthropic/claude-opus-5-5:high`.
|
|
472
|
+
- `task.enableEffort` stays at omp's default `false`, so the task tool has no
|
|
473
|
+
`effort` field and every spawn runs the level of its agent's role. When the
|
|
474
|
+
field is on, a caller can pass `effort: lo`, `med`, or `hi`, and caller
|
|
475
|
+
effort outranks the role suffix. omp maps it onto the model's own ladder,
|
|
476
|
+
not onto level names: `lo` selects the lowest level, `med` the lower middle,
|
|
477
|
+
and `hi` the highest (`resolveTaskEffortLevel` in omp's
|
|
478
|
+
`packages/tui/src/thinking.ts`, omp 18.8.5). Then omp caps the result at
|
|
479
|
+
`task.maxEffort`. On the `low, medium, high, xhigh, max` ladder, `hi` gave
|
|
480
|
+
`xhigh` under the `xhigh` cap, so many `task` spawns ran Opus 5.5 xhigh
|
|
481
|
+
despite the `:high` role suffix. On 2026-11-05, a fresh `omp -p` run on omp
|
|
482
|
+
18.16.0 with the field off had no `effort` field, and its `task` child
|
|
483
|
+
recorded `resolvedModel: anthropic/claude-opus-5-5:high`.
|
|
457
484
|
- `task.maxEffort` is `xhigh`. With `task.enableEffort` off, the cap applies
|
|
458
485
|
only to a per-call `model` with a `:level` suffix.
|
|
459
486
|
- In the deployed config, `scout` and `sonic` run GPT-6 Luna `medium`. The
|
|
@@ -475,10 +502,13 @@ AA's 1M.
|
|
|
475
502
|
`docks-kit omp [--model <selector>|--pick] [args...]` starts one interactive
|
|
476
503
|
omp session with a configuration overlay that selects a free model. The
|
|
477
504
|
overlay sets the 16 chat-model roles in `ompOverlay.ts ROLE_ORDER` to that
|
|
478
|
-
model and empties the
|
|
479
|
-
FALLBACK_ORDER`. These roles include `memory` and `
|
|
480
|
-
not set them, but omp otherwise uses its built-in priority chain for
|
|
481
|
-
and that chain can name a paid model.
|
|
505
|
+
model and empties the 12 retry fallback chains in `ompOverlay.ts
|
|
506
|
+
FALLBACK_ORDER`. These roles include `memory`, `judge`, and `advisor`. The SoT
|
|
507
|
+
does not set them, but omp otherwise uses its built-in priority chain for
|
|
508
|
+
`memory` and `judge`, and that chain can name a paid model. omp resolves the
|
|
509
|
+
`/advisor on` model from `modelRoles.advisor` before the `slow` role. A
|
|
510
|
+
deployed, user, or project `advisor` role could otherwise send it to a paid
|
|
511
|
+
model. The overlay also empties the retry
|
|
482
512
|
chains keyed by the free model: the bare selector, `<selector>:off`, and
|
|
483
513
|
`<selector>:<level>` for every level in `ompOverlay.ts THINKING_LADDER`. omp
|
|
484
514
|
uses a chain keyed by the model before a role chain, so a project or
|
|
@@ -491,16 +521,16 @@ stays in the merged config. A retry of the free model does not reach such a
|
|
|
491
521
|
chain, because omp uses the empty chain keyed by the free model first.
|
|
492
522
|
The overlay also sets `defaultThinkingLevel` and `task.maxEffort`, except for
|
|
493
523
|
a model that publishes no thinking ladder, where both keys are omitted and the
|
|
494
|
-
deployed values apply.
|
|
524
|
+
deployed or omp default values apply.
|
|
495
525
|
|
|
496
526
|
The launcher renders a run overlay to
|
|
497
527
|
`~/.cache/docks-kit/omp-free-<model>-<digest>.yml` at mode 0600 and passes it
|
|
498
528
|
through omp's repeatable `--config` flag. Each distinct model-and-level
|
|
499
529
|
combination gets its own file, because omp re-reads a changed overlay during
|
|
500
|
-
a live session, and a second launcher must not change the
|
|
530
|
+
a live session, and a second launcher must not change the level of that
|
|
501
531
|
session. The name sanitizes the selector for the file system and appends a
|
|
502
|
-
digest of the rendered overlay. The digest covers the selector and
|
|
503
|
-
|
|
532
|
+
digest of the rendered overlay. The digest covers the selector and the
|
|
533
|
+
level, so two selectors that differ only in separator characters stay
|
|
504
534
|
apart. A session launch never reads or writes `~/.omp/agent/config.yml` or
|
|
505
535
|
`models.yml`; only `--fast`, below, edits `config.yml`. It
|
|
506
536
|
never touches the SoT. The next plain `omp` run uses the paid configuration
|
|
@@ -576,11 +606,9 @@ Spark 1.3 Free", 1,048,576-token context) at `xhigh`. `xhigh` is the ceiling
|
|
|
576
606
|
of that model ladder: `minimal, low, medium, high, xhigh`. The paid
|
|
577
607
|
`muse-spark-1.3` sibling also offers `max`. The free variant does not.
|
|
578
608
|
|
|
579
|
-
|
|
580
|
-
|
|
581
|
-
|
|
582
|
-
review passes. When every level of a ladder is above `medium`, the advisor
|
|
583
|
-
takes the lowest level, so it never becomes the most expensive role.
|
|
609
|
+
The level comes from the ladder of the chosen model, never from a fixed list.
|
|
610
|
+
Every role, `advisor` included, runs the same level. Under `--model`, that
|
|
611
|
+
level is the ceiling of the ladder.
|
|
584
612
|
|
|
585
613
|
Free ladders are not uniform. On 2026-09-11 the 26 free models published six
|
|
586
614
|
distinct ladders: 21 include `medium`, 3 stop at `low, high, max` or
|
|
@@ -590,7 +618,7 @@ fail when the session starts.
|
|
|
590
618
|
|
|
591
619
|
The choice persists per machine in `~/.docks-kit/kit.db` (table
|
|
592
620
|
`omp_session`), next to the harness selection. It survives across sessions. It is never
|
|
593
|
-
committed. `--model <selector>` records one selector and derives
|
|
621
|
+
committed. `--model <selector>` records one selector and derives the level
|
|
594
622
|
from the catalog row. `--pick` opens an interactive wizard.
|
|
595
623
|
|
|
596
624
|
The picker lists only models the live `omp models --json` catalog reports at
|
|
@@ -598,31 +626,24 @@ zero input and output cost with `pricingStatus` `free` or `unknown` (26
|
|
|
598
626
|
entries on 2026-09-11). A zero-cost row with another status, such as
|
|
599
627
|
`variable` or `included`, is billed outside the token price, so the picker
|
|
600
628
|
leaves it out. The resulting overlay sets the 16 `ROLE_ORDER` roles to the
|
|
601
|
-
chosen free model and empties the
|
|
629
|
+
chosen free model and empties the 12 `FALLBACK_ORDER` chains and the chains
|
|
602
630
|
keyed by that model. The `web` role, its chain, and each other deployed chain
|
|
603
631
|
keep the deployed values. The
|
|
604
632
|
catalog can advertise a free model that the account cannot call; omp reports
|
|
605
633
|
that provider error unchanged.
|
|
606
634
|
|
|
607
|
-
After the model, the wizard asks about thinking
|
|
608
|
-
planEffortChoice` decides
|
|
635
|
+
After the model, the wizard asks about the thinking level. `ompOverlay.ts
|
|
636
|
+
planEffortChoice` decides whether the question applies, because omp accepts a
|
|
609
637
|
`:level` suffix only for a level the chosen model publishes:
|
|
610
638
|
|
|
611
639
|
- No ladder: the wizard asks nothing and the overlay omits every level.
|
|
612
640
|
- One level: the wizard states that level and asks nothing.
|
|
613
|
-
- Two or more levels: the wizard asks
|
|
614
|
-
|
|
615
|
-
|
|
616
|
-
|
|
617
|
-
The advisor question leads with a recommendation from `ompOverlay.ts
|
|
618
|
-
advisorRecommendation`: two steps down the model own ladder, clamped to the
|
|
619
|
-
lowest level that model publishes. The recommendation is the first row
|
|
620
|
-
because `Prompt.Select` starts on the first entry and accepts no initial
|
|
621
|
-
index, so `Enter` takes it. The full ladder follows in ladder order. A
|
|
622
|
-
recommendation equal to the chosen level is omitted rather than duplicated.
|
|
641
|
+
- Two or more levels: the wizard asks one question, the thinking level for
|
|
642
|
+
every role. It lists the levels of the model from the highest down, so
|
|
643
|
+
`Enter` keeps the highest level.
|
|
623
644
|
|
|
624
|
-
|
|
625
|
-
|
|
645
|
+
The launch line names the selector and the level, or the model default level
|
|
646
|
+
for a model without a ladder. It reports no separate advisor level.
|
|
626
647
|
|
|
627
648
|
An omp login is required. The kit owns no login flow. Every launch reads the
|
|
628
649
|
live free catalog of each candidate profile. A plain launch checks the
|
|
@@ -647,9 +668,9 @@ stops the launcher before omp starts.
|
|
|
647
668
|
|
|
648
669
|
When the catalog renames the free variant, change only the default selector
|
|
649
670
|
constant. Until then, a plain launch on that default stops with an error. A
|
|
650
|
-
ladder change needs no code change. Under `--model`
|
|
671
|
+
ladder change needs no code change. Under `--model` the level is derived
|
|
651
672
|
from the catalog row of the chosen model, and under `--pick` the user chooses
|
|
652
|
-
|
|
673
|
+
it from the ladder that same row publishes. Refresh the recorded default
|
|
653
674
|
and the ladder counts in these docs in the same commit.
|
|
654
675
|
|
|
655
676
|
## OpenAI fast mode (`docks-kit omp --fast`)
|
|
@@ -670,8 +691,8 @@ The four review agents inherit the live tier of the main session
|
|
|
670
691
|
and on standard processing when it is `none`. A running session keeps the tier
|
|
671
692
|
it started with, so its reviewers do not follow a later toggle.
|
|
672
693
|
|
|
673
|
-
`--fast` never changes the Anthropic tier
|
|
674
|
-
|
|
694
|
+
`--fast` never changes the Anthropic tier. The SoT leaves `tier.anthropic` at
|
|
695
|
+
omp's default `none`, so each session starts with it. In omp
|
|
675
696
|
18.16.0, `/fast on` in a session whose current
|
|
676
697
|
model is Anthropic sets `anthropic: priority` for that session. The reviewers
|
|
677
698
|
then inherit it, so their Opus 5.5 fallback also runs fast. omp has no
|
package/cli/src/commands/omp.ts
CHANGED
|
@@ -12,8 +12,6 @@ import {
|
|
|
12
12
|
type OmpSessionModel,
|
|
13
13
|
} from "../engine-native/harnesses";
|
|
14
14
|
import {
|
|
15
|
-
advisorLevelFor,
|
|
16
|
-
advisorRecommendation,
|
|
17
15
|
buildOmpArgs,
|
|
18
16
|
ladderCeiling,
|
|
19
17
|
ompChildEnvironment,
|
|
@@ -191,71 +189,39 @@ const loadFreeCatalog = (
|
|
|
191
189
|
});
|
|
192
190
|
|
|
193
191
|
/**
|
|
194
|
-
* Ask for the thinking
|
|
192
|
+
* Ask for the thinking level of one model. Every choice comes from that
|
|
195
193
|
* model own ladder, because omp rejects a `:level` suffix the model does not
|
|
196
194
|
* publish. A model with no ladder and a model with one level are settled
|
|
197
|
-
* without a question.
|
|
195
|
+
* without a question. Every role runs at the chosen level.
|
|
198
196
|
*/
|
|
199
|
-
const
|
|
197
|
+
const pickLevel = (catalogModel: CatalogModel) =>
|
|
200
198
|
Effect.gen(function* () {
|
|
201
199
|
const plan = planEffortChoice(catalogModel.thinking);
|
|
202
200
|
if (plan.kind === "none") {
|
|
203
201
|
yield* Console.error(
|
|
204
202
|
`${catalogModel.name} publishes no thinking levels; the model default applies`,
|
|
205
203
|
);
|
|
206
|
-
return
|
|
204
|
+
return undefined;
|
|
207
205
|
}
|
|
208
206
|
if (plan.kind === "fixed") {
|
|
209
207
|
yield* Console.error(`${catalogModel.name} publishes one thinking level: ${plan.level}`);
|
|
210
|
-
return
|
|
208
|
+
return plan.level;
|
|
211
209
|
}
|
|
212
|
-
|
|
213
|
-
|
|
214
|
-
|
|
215
|
-
|
|
216
|
-
|
|
217
|
-
|
|
218
|
-
|
|
219
|
-
|
|
220
|
-
|
|
221
|
-
|
|
210
|
+
// Prompt.Select always starts on the first entry and offers no initial
|
|
211
|
+
// index, so the list runs highest first and Enter keeps the ceiling.
|
|
212
|
+
return yield* Prompt.Select({
|
|
213
|
+
message: "Thinking level for every role",
|
|
214
|
+
choices: [...plan.levels].reverse().map((level) => ({
|
|
215
|
+
title: level,
|
|
216
|
+
value: level,
|
|
217
|
+
description:
|
|
218
|
+
level === plan.highest
|
|
219
|
+
? "highest this model offers"
|
|
220
|
+
: level === plan.levels[0]
|
|
221
|
+
? "lowest this model offers"
|
|
222
|
+
: "",
|
|
223
|
+
})),
|
|
222
224
|
});
|
|
223
|
-
if (uniform) return { thinking: plan.highest, advisorThinking: plan.highest };
|
|
224
|
-
const ladderChoices = plan.levels.map((level) => ({
|
|
225
|
-
title: level,
|
|
226
|
-
value: level,
|
|
227
|
-
description:
|
|
228
|
-
level === plan.highest
|
|
229
|
-
? "highest this model offers"
|
|
230
|
-
: level === plan.levels[0]
|
|
231
|
-
? "lowest this model offers"
|
|
232
|
-
: "",
|
|
233
|
-
}));
|
|
234
|
-
const thinking = yield* Prompt.Select({
|
|
235
|
-
message: "Thinking level for all roles except the advisor",
|
|
236
|
-
choices: ladderChoices,
|
|
237
|
-
});
|
|
238
|
-
const suggested = advisorRecommendation(plan.levels, thinking);
|
|
239
|
-
// The list leads with the suggestion, because Prompt.Select always starts
|
|
240
|
-
// on the first entry and offers no initial index. The full ladder follows
|
|
241
|
-
// in ladder order, so a different level stays one keypress away. A
|
|
242
|
-
// suggestion equal to the main level would only duplicate a row.
|
|
243
|
-
const advisorChoices =
|
|
244
|
-
suggested === undefined || suggested === thinking
|
|
245
|
-
? ladderChoices
|
|
246
|
-
: [
|
|
247
|
-
{
|
|
248
|
-
title: suggested,
|
|
249
|
-
value: suggested,
|
|
250
|
-
description: `recommended, two levels below ${thinking}`,
|
|
251
|
-
},
|
|
252
|
-
...ladderChoices,
|
|
253
|
-
];
|
|
254
|
-
const advisorThinking = yield* Prompt.Select({
|
|
255
|
-
message: "Thinking level for the advisor role",
|
|
256
|
-
choices: advisorChoices,
|
|
257
|
-
});
|
|
258
|
-
return { thinking, advisorThinking };
|
|
259
225
|
});
|
|
260
226
|
|
|
261
227
|
/**
|
|
@@ -364,7 +330,6 @@ export const ompCommand = Command.make("omp", { model, pick, fast, args }, (conf
|
|
|
364
330
|
chosen = {
|
|
365
331
|
selector: resolved.model.selector,
|
|
366
332
|
thinking: ladderCeiling(resolved.model.thinking),
|
|
367
|
-
advisorThinking: advisorLevelFor(resolved.model.thinking),
|
|
368
333
|
};
|
|
369
334
|
}
|
|
370
335
|
|
|
@@ -387,12 +352,7 @@ export const ompCommand = Command.make("omp", { model, pick, fast, args }, (conf
|
|
|
387
352
|
if (match === undefined) {
|
|
388
353
|
return yield* bail(`Model '${pickedSelector}' is not in the free catalog`);
|
|
389
354
|
}
|
|
390
|
-
|
|
391
|
-
chosen = {
|
|
392
|
-
selector: match.selector,
|
|
393
|
-
thinking: chosenLevels.thinking,
|
|
394
|
-
advisorThinking: chosenLevels.advisorThinking,
|
|
395
|
-
};
|
|
355
|
+
chosen = { selector: match.selector, thinking: yield* pickLevel(match) };
|
|
396
356
|
}
|
|
397
357
|
|
|
398
358
|
const session: OmpSessionModel =
|
|
@@ -427,21 +387,13 @@ export const ompCommand = Command.make("omp", { model, pick, fast, args }, (conf
|
|
|
427
387
|
});
|
|
428
388
|
|
|
429
389
|
// A level-free model states the model default so the line never shows
|
|
430
|
-
// an undefined level.
|
|
431
|
-
// a uniform session has nothing extra to report.
|
|
432
|
-
const level =
|
|
433
|
-
typeof session.thinking === "string" && session.thinking.trim() !== ""
|
|
434
|
-
? session.thinking
|
|
435
|
-
: undefined;
|
|
436
|
-
const advisor =
|
|
437
|
-
typeof session.advisorThinking === "string" && session.advisorThinking.trim() !== ""
|
|
438
|
-
? session.advisorThinking
|
|
439
|
-
: undefined;
|
|
390
|
+
// an undefined level.
|
|
440
391
|
const levelText =
|
|
441
|
-
|
|
442
|
-
|
|
392
|
+
typeof session.thinking === "string" && session.thinking.trim() !== ""
|
|
393
|
+
? `${session.thinking} thinking`
|
|
394
|
+
: "the model default thinking level";
|
|
443
395
|
yield* Console.error(
|
|
444
|
-
`Starting omp with ${session.selector} at ${levelText}
|
|
396
|
+
`Starting omp with ${session.selector} at ${levelText} (session only; deployed config unchanged)`,
|
|
445
397
|
);
|
|
446
398
|
|
|
447
399
|
// An exported model variable would set its role above the overlay, so the
|
|
@@ -18,7 +18,6 @@ export type { OmpSessionModel };
|
|
|
18
18
|
export const DEFAULT_OMP_SESSION_MODEL: OmpSessionModel = {
|
|
19
19
|
selector: "opencode-zen/muse-spark-1.3-contributor-free",
|
|
20
20
|
thinking: "xhigh",
|
|
21
|
-
advisorThinking: "medium",
|
|
22
21
|
};
|
|
23
22
|
|
|
24
23
|
function isHarness(value: unknown): value is Harness {
|
|
@@ -104,17 +103,13 @@ export function writeHarnessSelection(home: string, selection: ReadonlyArray<Har
|
|
|
104
103
|
export function readOmpSessionModel(home: string): OmpSessionModel | undefined {
|
|
105
104
|
try {
|
|
106
105
|
return withKitDb(home, "read", (db) => {
|
|
107
|
-
const row = db
|
|
108
|
-
.prepare("SELECT selector, thinking, advisor_thinking FROM omp_session WHERE id = 1")
|
|
109
|
-
.get();
|
|
106
|
+
const row = db.prepare("SELECT selector, thinking FROM omp_session WHERE id = 1").get();
|
|
110
107
|
const selector = row?.["selector"];
|
|
111
108
|
if (!isNonBlankString(selector)) return undefined;
|
|
112
109
|
const thinking = row?.["thinking"];
|
|
113
|
-
const advisorThinking = row?.["advisor_thinking"];
|
|
114
110
|
const model: OmpSessionModel = {
|
|
115
111
|
selector,
|
|
116
112
|
...(isNonBlankString(thinking) ? { thinking } : {}),
|
|
117
|
-
...(isNonBlankString(advisorThinking) ? { advisorThinking } : {}),
|
|
118
113
|
};
|
|
119
114
|
return model;
|
|
120
115
|
});
|
|
@@ -129,16 +124,15 @@ export function writeOmpSessionModel(home: string, model: OmpSessionModel): void
|
|
|
129
124
|
if (!isNonBlankString(model.selector)) {
|
|
130
125
|
throw new Error("Omp session model selector must be a non-empty string");
|
|
131
126
|
}
|
|
132
|
-
// Store blank or absent
|
|
127
|
+
// Store a blank or absent level as NULL so a switch to a level-free model
|
|
133
128
|
// leaves no stale level behind.
|
|
134
129
|
const thinking = nonBlankOrNull(model.thinking);
|
|
135
|
-
const advisorThinking = nonBlankOrNull(model.advisorThinking);
|
|
136
130
|
withKitDb(home, "write", (db) =>
|
|
137
131
|
db
|
|
138
132
|
.prepare(
|
|
139
|
-
`INSERT INTO omp_session (id, selector, thinking
|
|
140
|
-
ON CONFLICT(id) DO UPDATE SET selector = excluded.selector, thinking = excluded.thinking
|
|
133
|
+
`INSERT INTO omp_session (id, selector, thinking) VALUES (1, ?, ?)
|
|
134
|
+
ON CONFLICT(id) DO UPDATE SET selector = excluded.selector, thinking = excluded.thinking`,
|
|
141
135
|
)
|
|
142
|
-
.run(model.selector, thinking
|
|
136
|
+
.run(model.selector, thinking),
|
|
143
137
|
);
|
|
144
138
|
}
|