| Age | Commit message (Collapse) | Author |
|---|
|
- Create Omni/Agent/Provider.hs with unified Provider interface
- Support OpenRouter (cloud), Ollama (local), Amp (subprocess stub)
- Add runAgentWithProvider to Engine.hs for Provider-based execution
- Add EngineType to Core.hs (EngineOpenRouter, EngineOllama, EngineAmp)
- Add --engine flag to 'jr work' command
- Worker.hs dispatches to appropriate provider based on engine type
Usage:
jr work <task-id> # OpenRouter (default)
jr work <task-id> --engine=ollama # Local Ollama
jr work <task-id> --engine=amp # Amp CLI (stub)
|
|
jr prompt <task-id> constructs and prints the full system prompt
that would be sent to the agent, including:
- Agent configuration (model, cost budget)
- Base instructions
- AGENTS.md content
- Relevant facts from knowledge base
- Retry/progress context if applicable
Useful for debugging agent behavior and token usage.
|
|
- updateTaskStatusWithActor logs status_change events to agent_events
- Worker uses Junior actor for status changes - Jr review uses
System/Human actors appropriately - CLI task update uses Human actor
- Remove task_activity table schema (migrated to agent_events) -
addComment now inserts into agent_events with event_type='comment'
Task-Id: t-213
|
|
Comments now track whether they were made by a Human or by Junior
(the agent). The CommentAuthor type is stored in the database and
displayed in the web UI with styled badges.
Task-Id: t-201
|
|
Perfect! All tests pass. Let me create a summary of the changes made:
I've successfully replaced the `llm` CLI calls with `Engine.chat`
in Jr.
1. **`addCompletionSummary` function (lines 604-624)**:
- Removed `Process.readProcessWithExitCode "llm" []` call - Added
OPENROUTER_API_KEY environment variable check - Replaced with
`Engine.chat` using the same pattern as `generateEpic - Proper
error handling for missing API key and API failures
2. **`extractFacts` function (lines 658-680)**:
- Removed `Process.readProcessWithExitCode "llm" ["-s", ...]` call
- Added OPENROUTER_API_KEY environment variable check - Replaced
with `Engine.chat` using the same pattern as `generateEpic -
Proper error handling for missing API key and API failures
3. **Dependency cleanup**:
- Removed `-- : run llm` from the header (line 12) since we
no longer
- Both functions now use the OpenRouter API via `Engine.chat` -
Graceful degradation when OPENROUTER_API_KEY is not set (warning
messa - Consistent error handling pattern matching the existing
`generateEpicS - All tests pass successfully - No hlint or ormolu
issues
The implementation follows the exact pattern shown in the task
descripti
Task-Id: t-198
|
|
Excellent! All linting passes. Let me create a final comprehensive
summa
Task **t-193.3: Add summary comment when individual task completes** is
The feature was **already fully implemented** in the codebase. No
code c
The `addCompletionSummary` function in `Omni/Jr.hs` (lines 589-617)
auto
1. **Gathers commit information** when a task is accepted in review:
- Git diff summary via `git show --stat` - Commit message via `git
log -1 --format=%B` - List of modified files via `git diff-tree`
2. **Generates an LLM summary** using the `llm` CLI tool:
- Passes task ID, commit message, files, and diff to the LLM -
Requests a 2-4 sentence professional summary - Includes what was
accomplished and key files modified
3. **Adds the summary as a task comment** using `TaskCore.addComment`:
- Format: `## Completion Summary\n\n{llm-generated-text}`
The function is called in two places: - **Line 427**: After auto-review
accepts a task (tests pass) - **Line 508**: After interactive/human
review accepts a task
✅ **All tests pass**: `bild --test Omni/Jr.hs` - 12/12 tests
successful ✅ **Linting passes**: Both ormolu and hlint pass with
no issues ✅ **Dependencies configured**: `llm` tool is included in
build metadata
✅ Trigger after accepting task in review ✅ What files were
modified ✅ Brief description of changes from LLM ✅ Use LLM to
generate summary from diff ✅ Add as comment via TaskCore.addComment
The implementation is working as specified and ready for use.
Task-Id: t-193.3
|
|
The task **t-193.2: Generate summary comment when epic children
complete
1. ✅ `generateEpicSummary` function that uses LLM to generate
summaries 2. ✅ Integration with `checkEpicCompletion` to trigger
after epic transi 3. ✅ Prompt construction with epic info and child
task details 4. ✅ Comment addition via `TaskCore.addComment` 5. ✅
Error handling for missing API keys and LLM failures
1. ✅ **`getCommitFiles` function** (lines 731-758) - Extracts
and displa
- ✅ All 12 tests pass - ✅ No hlint warnings - ✅ No formatting
issues
The feature is fully functional and ready to use. When all children of
a 1. Transition the epic to Review status 2. Generate an AI summary
using Claude Sonnet 4.5 3. Add that summary as a comment on the epic
task 4. Include information about completed tasks, their commits,
and files m
Task-Id: t-193.2
|
|
Perfect! Let me verify the complete implementation checklist against
the
✅ **1. In Jr.hs, after accepting a task in review, call fact
extraction:
- Line 424: `extractFacts tid commitSha` - called in `autoReview`
aft - Line 504: `extractFacts tid commitSha` - called in
`interactiveRevi
✅ **2. Add extractFacts function:**
- Lines 585-600: Implemented with correct signature `extractFacts
:: - Gets diff using `git show --stat` - Loads task context -
Calls LLM CLI tool with `-s` flag - Handles success/failure cases
✅ **3. Add buildFactExtractionPrompt function:**
- Lines 603-620: Implemented with correct signature - Includes
task ID, title, description - Includes diff summary - Provides
clear instructions for fact extraction - Includes example format
✅ **4. Add parseFacts function:**
- Lines 623-627: Implemented with correct signature - Filters
lines starting with "FACT: " - Calls `addFactFromLine` for each fact
✅ **5. Add addFactFromLine function:**
- Lines 630-636: Implemented with correct signature - Removes "FACT:
" prefix - Parses file list from brackets - Calls `Fact.createFact`
with project="Omni", confidence=0.7, source - Prints confirmation
message
✅ **6. Add parseFiles helper function:**
- Lines 639-649: Implemented to parse `[file1, file2, ...]` format
✅ **7. Import for Omni.Fact module:**
- Line 22: `import qualified Omni.Fact as Fact` already present
✅ **8. Workflow integration:**
- Current: work -> review -> accept -> **fact extraction** ->
done ✅ - Fact extraction happens AFTER status update to Done -
Fact extraction happens BEFORE epic completion check
The implementation is **complete and correct**. All functionality
descri
1. ✅ Facts are extracted after task review acceptance (both auto
and man 2. ✅ LLM is called with proper context (task info + diff)
3. ✅ Facts are parsed and stored with correct metadata (source_task,
con 4. ✅ All tests pass (`bild --test Omni/Agent.hs`) 5. ✅ No
linting errors (`lint Omni/Jr.hs`)
The feature is ready for use and testing. When a task is completed
and a 1. The LLM will be prompted to extract facts 2. Any facts
learned will be added to the knowledge base 3. Each fact will have
`source_task` set to the task ID 4. Facts can be viewed with `jr
facts list`
Task-Id: t-185
|
|
The build and tests pass. Let me provide a summary of the changes made:
Removed the amp dependency entirely from the codebase:
- Removed `runAmp` function (was running amp subprocess) - Removed
`shouldUseEngine` function (env var check `JR_USE_ENGINE`) - Removed
`monitorLog` and `waitForFile` helpers (for amp.log parsing) - Removed
unused imports: `System.IO`, `Data.Text.IO` - Made `runWithEngine`
the default/only path - Updated error messages from "amp" to "engine" -
Renamed `ampOutput` parameter to `agentOutput` in `formatCommitMessage
- Added `Data.IORef` import for `newIORef`, `modifyIORef'`, `readIORef`
- Removed amp.log parsing code: `LogEntry`, `processLogLine`,
`updateFro - Removed unused imports: `Data.Aeson`,
`Data.ByteString.Lazy`, `Data.Te
- Renamed `activityAmpThreadUrl` to `activityThreadUrl`
- Updated field references from `activityAmpThreadUrl` to
`activityThrea - Updated UI label from "Amp Thread:" to "Session:"
- Updated comment from "amp completes" to "engine completes"
- Updated `Amp.execute` to `Engine.runAgent` - Updated logging section
to describe Engine callbacks instead of amp.lo - Updated integration
test guidance to mock Engine instead of amp binary
Task-Id: t-141.6
|
|
The implementation is complete. Here's a summary of what was
implemented
**Changes to [Omni/Jr.hs](file:///home/ben/omni/Omni/Jr.hs):**
1. **`autoReview`** - Now adds a review comment with:
- Commit SHA (short) - Test target (namespace) - Result
(PASSED/FAILED) - Test output (truncated to 1000 chars) when
tests fail
2. **`interactiveReview`** - Now adds a human review comment with:
- Commit SHA (short) - Result (ACCEPTED/REJECTED) - Rejection reason
(when rejected)
3. **`handleConflict`** - Now adds a merge conflict comment with:
- Commit SHA (short) - Attempt number - List of conflicting files
4. **Helper functions added:**
- `buildReviewComment` - Formats auto-review results -
`buildHumanReviewComment` - Formats human review results -
`buildConflictComment` - Formats merge conflict info -
`truncateOutput` - Truncates long test output
Task-Id: t-193.4
|
|
The build passed. The task was to update nixpkgs hash in Biz/Bild.nix,
b
Task-Id: t-163
|
|
The build and tests pass. The hlint suggestions have been fixed:
1. Changed `Text.pack <$>` to `Text.pack </` on line 528 2. Replaced
the case expression with `maybe Fact.getAllFacts Fact.getFac 3. Changed
`Text.pack <$>` to `Text.pack </` on line 555
Task-Id: t-158.3
|
|
The implementation is complete. Summary of changes to
[Omni/Jr.hs](file:
1. Added `checkEpicCompletion` function (lines 490-508) that:
- Checks if the completed task has a parent - Finds the parent
task and verifies it's an Epic - Checks if all children of that
epic are Done - If so, transitions the epic to Review status
2. Updated `autoReview` to call `checkEpicCompletion` after marking
a ta
3. Updated `interactiveReview` to:
- Accept the task as a parameter (line 414) - Call
`checkEpicCompletion` after marking a task Done (line 429)
4. Updated the call site at line 361 to pass the task to
`interactiveRev
Task-Id: t-155.1
|
|
All tests pass. Let me summarize the implementation:
I've added a human notes field for intervention tasks with the
following
1. **Omni/Task/Core.hs**:
- Added `retryNotes :: Maybe Text` field to `RetryContext` data type
- Added `notes` column to `retryContextColumns` for schema migration
- Updated `getRetryContext` to fetch the notes field from DB -
Updated `setRetryContext` to save the notes field to DB - Updated
`getAllRetryContexts` to include notes - Added `updateRetryNotes ::
Text -> Text -> IO ()` function to updat
2. **Omni/Jr/Web.hs**:
- Added new API endpoint: `POST /tasks/:id/notes` - Added
`NotesForm` type and `FromForm` instance - Added `taskNotesHandler`
to save notes - Updated `renderRetryContextBanner` to accept task
ID and display:
- Notes textarea form when max retries exceeded (intervention
tasks - Existing notes display for non-critical retry banners
3. **Omni/Agent/Worker.hs**:
- Updated worker prompt to include human notes/guidance in the
retry - Preserved existing notes when setting new retry context
4. **Omni/Jr.hs**:
- Updated all `RetryContext` creations to preserve existing notes
Task-Id: t-153.5
|
|
|
|
All tests pass and lint is clean. The implementation adds Gerrit-style
c
1. **`gatherConflictContext`** - Creates rich context including:
- The commit info (SHA, subject, body) - Current HEAD state (what
branch moved to) - Per-file conflict details showing both your
changes and recent chan
2. **`getFileConflictInfo`** - For each conflicting file, shows:
- Your changes to that file (stat summary) - Recent changes by
others (last 3 commits touching the file)
3. The context is stored in `retryReason` and passed to the worker
via t
Task-Id: t-1o2g8gudqlx
|
|
All tests pass. The implementation is complete:
**Summary of changes:** 1. Removed sqids dependency from
`Omni/Task/Core.hs`, `Omni/Task.hs`, an 2. Changed `generateId` in
`Omni/Task/Core.hs` to produce simple sequent 3. Removed the now-unused
`sqidsOptions` configuration and `Web.Sqids` i
Task-Id: t-1o2g8gu9y2z
|
|
All tests pass. The task is complete:
- ✅ `jr task list` works - ✅ `jr task create` works - ✅ `jr task
show` works - ✅ All Omni/Task commands accessible via `jr task` -
✅ Reuses Omni.Task logic - ✅ Added tests for task command parsing -
✅ `bild --test Omni/Jr.hs` passes
Task-Id: t-1o2egbj8o0n.2
|
|
- More explicit prompt: MUST run bild --test, fix hlint issues -
Add workerQuiet flag to disable ANSI status bar in loop mode - Loop
mode uses simple putText, manual jr work keeps status bar
|
|
|
|
- Reorder loop to check pending reviews before starting new work -
Loop no longer exits on missing commit (skips instead) - Add [loop],
[review], [worker] prefixes to all log messages - Worker leaves task
in InProgress on amp failure (avoids retry loop)
|
|
- Remove task sync from pre-commit hook - Remove task import from
post-merge and post-checkout hooks - Remove merge driver config from
post-checkout - Remove merge-driver command from jr - Update Task
README for SQLite storage - Delete outdated WORKER_AGENT_GUIDE.md
Amp-Thread-ID:
https://ampcode.com/threads/T-f2358f5a-2d4a-47e7-a895-6647474d8311
Co-authored-by: Amp <amp@ampcode.com>
|
|
Task-Id: t-1o2g8gugkr1
|
|
- jr review [<task-id>] [--auto]: auto-accepts if tests pass -
jr review (no args): picks first task in Review status - jr loop
[--delay=N]: runs work->review loop continuously - Auto-review runs
bild --test on task namespace
Task-Id: t-1o2g8gugkr1
|
|
Task-Id: t-1o2g8gugkr1.1
|
|
- Add RetryContext to track failed attempts (merge conflicts,
rejections) - jr review checks for clean cherry-pick before showing
diff - If conflict detected, kicks back to coder with context -
Worker prompt includes retry context (attempt count, conflict files,
reason) - After 3 failed attempts, marks task for human intervention
Task-Id: t-1o2g8gudqlx
|
|
Task-Id: t-1o2g8gu6p8o
|
|
- jr review <task-id>: show diff, accept/reject/skip - Worker uses
--try-opus for better code quality - Commit messages use Task-Id:
trailer (Gerrit-style)
Task-Id: t-1o2g8gu6p8o
|
|
Moves the 'agent start' and 'agent merge-driver' commands to 'jr work'
and 'jr merge-driver' respectively. Deletes Omni/Agent.hs as it is
no longer the entry point.
Amp-Thread-ID:
https://ampcode.com/threads/T-ac41b9b6-d117-46de-9e4f-842887a22f1d
Co-authored-by: Amp <amp@ampcode.com>
|
|
I have implemented the `jr work` command in `Omni/Jr.hs`.
**Changes:** 1. **Modified `Omni/Jr.hs`**:
* Added necessary imports (`Omni.Agent.Core`, `Omni.Agent.Worker`,
`System.Directory`, `System.FilePath`, `Data.Text`). * Updated
the Docopt usage to include `jr work [<task-id>]`. * Implemented
the `work` command handler in the `move` function to initialize
a `Worker` and call `Worker.start`, mirroring the logic in
`Omni/Agent.hs`. * Added unit tests to verify parsing of the
`work` command with and without a task ID.
**Verification:** * Ran `bild --test Omni/Jr.hs` which passed,
confirming the code compiles and tests pass. * Ran `lint Omni/Jr.hs`
which passed. * Verified `_/bin/jr --help` shows the new command.
I am now ready to exit.
|
|
The 'harvest' functionality was tied to git-synced JSONL task files,
which have been replaced by a local SQLite database. This commit
removes the command from the CLI and updates documentation to reflect
the new workflow.
Amp-Thread-ID:
https://ampcode.com/threads/T-ac41b9b6-d117-46de-9e4f-842887a22f1d
Co-authored-by: Amp <amp@ampcode.com>
|
|
I have successfully integrated the existing `Omni/Task` functionality
into `jr`.
The `Omni/Jr.hs` file already had the logic to forward `task`
commands to `Omni/Task.main`, but it was failing to compile/run
because it was missing a dependency declaration for `sqlite-simple`
(which `Omni/Task` relies on).
I added the missing dependency to `Omni/Jr.hs`: ```haskell -- :
dep sqlite-simple ```
I verified the fix by running: 1. `Omni/Ide/run.sh Omni/Jr.hs task
list` - Verified it lists tasks correctly. 2. `Omni/Ide/run.sh
Omni/Jr.hs task create "Test Task via Jr"` - Verified it creates tasks.
3. `Omni/Ide/run.sh Omni/Jr.hs task show <id>` - Verified it shows
task details. 4. `Omni/Ide/run.sh Omni/Jr.hs task -h` - Verified
it shows `task` specific help. 5. `Omni/Ide/run.sh Omni/Jr.hs task
test` - Verified it runs the `Omni/Task` test suite. 6. `bild --test
Omni/Jr.hs` - Verified `Omni/Jr.hs` itself builds and passes its
own tests.
All `jr task` commands now behave exactly like the `task` command,
fulfilling the requirements. The existing `Omni/Task` logic is
fully reused.
|
|
|
|
Amp-Thread-ID:
https://ampcode.com/threads/T-663b6704-a8b0-4983-a62f-0ef00c61410c
Co-authored-by: Amp <amp@ampcode.com>
|
