Metadata-Version: 2.4
Name: oddish
Version: 0.1.1
Summary: Postgres-backed eval scheduler for Harbor agent tasks — queuing, retries, and monitoring
Project-URL: Homepage, https://github.com/abundant-ai/oddish
Project-URL: Repository, https://github.com/abundant-ai/oddish
Project-URL: Issues, https://github.com/abundant-ai/oddish/issues
Project-URL: Documentation, https://github.com/abundant-ai/oddish/blob/main/oddish/README.md
Author-email: Rishi Desai <rishi@abundant.ai>
License:                                  Apache License
                                   Version 2.0, January 2004
                                http://www.apache.org/licenses/
        
           TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
        
           1. Definitions.
        
              "License" shall mean the terms and conditions for use, reproduction,
              and distribution as defined by Sections 1 through 9 of this document.
        
              "Licensor" shall mean the copyright owner or entity authorized by
              the copyright owner that is granting the License.
        
              "Legal Entity" shall mean the union of the acting entity and all
              other entities that control, are controlled by, or are under common
              control with that entity. For the purposes of this definition,
              "control" means (i) the power, direct or indirect, to cause the
              direction or management of such entity, whether by contract or
              otherwise, or (ii) ownership of fifty percent (50%) or more of the
              outstanding shares, or (iii) beneficial ownership of such entity.
        
              "You" (or "Your") shall mean an individual or Legal Entity
              exercising permissions granted by this License.
        
              "Source" form shall mean the preferred form for making modifications,
              including but not limited to software source code, documentation
              source, and configuration files.
        
              "Object" form shall mean any form resulting from mechanical
              transformation or translation of a Source form, including but
              not limited to compiled object code, generated documentation,
              and conversions to other media types.
        
              "Work" shall mean the work of authorship, whether in Source or
              Object form, made available under the License, as indicated by a
              copyright notice that is included in or attached to the work
              (an example is provided in the Appendix below).
        
              "Derivative Works" shall mean any work, whether in Source or Object
              form, that is based on (or derived from) the Work and for which the
              editorial revisions, annotations, elaborations, or other modifications
              represent, as a whole, an original work of authorship. For the purposes
              of this License, Derivative Works shall not include works that remain
              separable from, or merely link (or bind by name) to the interfaces of,
              the Work and Derivative Works thereof.
        
              "Contribution" shall mean any work of authorship, including
              the original version of the Work and any modifications or additions
              to that Work or Derivative Works thereof, that is intentionally
              submitted to Licensor for inclusion in the Work by the copyright owner
              or by an individual or Legal Entity authorized to submit on behalf of
              the copyright owner. For the purposes of this definition, "submitted"
              means any form of electronic, verbal, or written communication sent
              to the Licensor or its representatives, including but not limited to
              communication on electronic mailing lists, source code control systems,
              and issue tracking systems that are managed by, or on behalf of, the
              Licensor for the purpose of discussing and improving the Work, but
              excluding communication that is conspicuously marked or otherwise
              designated in writing by the copyright owner as "Not a Contribution."
        
              "Contributor" shall mean Licensor and any individual or Legal Entity
              on behalf of whom a Contribution has been received by Licensor and
              subsequently incorporated within the Work.
        
           2. Grant of Copyright License. Subject to the terms and conditions of
              this License, each Contributor hereby grants to You a perpetual,
              worldwide, non-exclusive, no-charge, royalty-free, irrevocable
              copyright license to reproduce, prepare Derivative Works of,
              publicly display, publicly perform, sublicense, and distribute the
              Work and such Derivative Works in Source or Object form.
        
           3. Grant of Patent License. Subject to the terms and conditions of
              this License, each Contributor hereby grants to You a perpetual,
              worldwide, non-exclusive, no-charge, royalty-free, irrevocable
              (except as stated in this section) patent license to make, have made,
              use, offer to sell, sell, import, and otherwise transfer the Work,
              where such license applies only to those patent claims licensable
              by such Contributor that are necessarily infringed by their
              Contribution(s) alone or by combination of their Contribution(s)
              with the Work to which such Contribution(s) was submitted. If You
              institute patent litigation against any entity (including a
              cross-claim or counterclaim in a lawsuit) alleging that the Work
              or a Contribution incorporated within the Work constitutes direct
              or contributory patent infringement, then any patent licenses
              granted to You under this License for that Work shall terminate
              as of the date such litigation is filed.
        
           4. Redistribution. You may reproduce and distribute copies of the
              Work or Derivative Works thereof in any medium, with or without
              modifications, and in Source or Object form, provided that You
              meet the following conditions:
        
              (a) You must give any other recipients of the Work or
                  Derivative Works a copy of this License; and
        
              (b) You must cause any modified files to carry prominent notices
                  stating that You changed the files; and
        
              (c) You must retain, in the Source form of any Derivative Works
                  that You distribute, all copyright, patent, trademark, and
                  attribution notices from the Source form of the Work,
                  excluding those notices that do not pertain to any part of
                  the Derivative Works; and
        
              (d) If the Work includes a "NOTICE" text file as part of its
                  distribution, then any Derivative Works that You distribute must
                  include a readable copy of the attribution notices contained
                  within such NOTICE file, excluding those notices that do not
                  pertain to any part of the Derivative Works, in at least one
                  of the following places: within a NOTICE text file distributed
                  as part of the Derivative Works; within the Source form or
                  documentation, if provided along with the Derivative Works; or,
                  within a display generated by the Derivative Works, if and
                  wherever such third-party notices normally appear. The contents
                  of the NOTICE file are for informational purposes only and
                  do not modify the License. You may add Your own attribution
                  notices within Derivative Works that You distribute, alongside
                  or as an addendum to the NOTICE text from the Work, provided
                  that such additional attribution notices cannot be construed
                  as modifying the License.
        
              You may add Your own copyright statement to Your modifications and
              may provide additional or different license terms and conditions
              for use, reproduction, or distribution of Your modifications, or
              for any such Derivative Works as a whole, provided Your use,
              reproduction, and distribution of the Work otherwise complies with
              the conditions stated in this License.
        
           5. Submission of Contributions. Unless You explicitly state otherwise,
              any Contribution intentionally submitted for inclusion in the Work
              by You to the Licensor shall be under the terms and conditions of
              this License, without any additional terms or conditions.
              Notwithstanding the above, nothing herein shall supersede or modify
              the terms of any separate license agreement you may have executed
              with Licensor regarding such Contributions.
        
           6. Trademarks. This License does not grant permission to use the trade
              names, trademarks, service marks, or product names of the Licensor,
              except as required for reasonable and customary use in describing the
              origin of the Work and reproducing the content of the NOTICE file.
        
           7. Disclaimer of Warranty. Unless required by applicable law or
              agreed to in writing, Licensor provides the Work (and each
              Contributor provides its Contributions) on an "AS IS" BASIS,
              WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
              implied, including, without limitation, any warranties or conditions
              of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
              PARTICULAR PURPOSE. You are solely responsible for determining the
              appropriateness of using or redistributing the Work and assume any
              risks associated with Your exercise of permissions under this License.
        
           8. Limitation of Liability. In no event and under no legal theory,
              whether in tort (including negligence), contract, or otherwise,
              unless required by applicable law (such as deliberate and grossly
              negligent acts) or agreed to in writing, shall any Contributor be
              liable to You for damages, including any direct, indirect, special,
              incidental, or consequential damages of any character arising as a
              result of this License or out of the use or inability to use the
              Work (including but not limited to damages for loss of goodwill,
              work stoppage, computer failure or malfunction, or any and all
              other commercial damages or losses), even if such Contributor
              has been advised of the possibility of such damages.
        
           9. Accepting Warranty or Additional Liability. While redistributing
              the Work or Derivative Works thereof, You may choose to offer,
              and charge a fee for, acceptance of support, warranty, indemnity,
              or other liability obligations and/or rights consistent with this
              License. However, in accepting such obligations, You may act only
              on Your own behalf and on Your sole responsibility, not on behalf
              of any other Contributor, and only if You agree to indemnify,
              defend, and hold each Contributor harmless for any liability
              incurred by, or claims asserted against, such Contributor by reason
              of your accepting any such warranty or additional liability.
        
           END OF TERMS AND CONDITIONS
        
           APPENDIX: How to apply the Apache License to your work.
        
              To apply the Apache License to your work, attach the following
              boilerplate notice, with the fields enclosed by brackets "[]"
              replaced with your own identifying information. (Don't include
              the brackets!)  The text should be enclosed in the appropriate
              comment syntax for the file format. We also recommend that a
              file or class name and description of purpose be included on the
              same "printed page" as the copyright notice for easier
              identification within third-party archives.
        
           Copyright [yyyy] [name of copyright owner]
        
           Licensed under the Apache License, Version 2.0 (the "License");
           you may not use this file except in compliance with the License.
           You may obtain a copy of the License at
        
               http://www.apache.org/licenses/LICENSE-2.0
        
           Unless required by applicable law or agreed to in writing, software
           distributed under the License is distributed on an "AS IS" BASIS,
           WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
           See the License for the specific language governing permissions and
           limitations under the License.
        
License-File: LICENSE
Keywords: agent,benchmark,eval,harbor,llm,scheduler
Classifier: Development Status :: 4 - Beta
Classifier: Intended Audience :: Developers
Classifier: License :: OSI Approved :: Apache Software License
Classifier: Programming Language :: Python :: 3
Classifier: Programming Language :: Python :: 3.12
Classifier: Programming Language :: Python :: 3.13
Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
Classifier: Topic :: Software Development :: Testing
Requires-Python: >=3.12
Requires-Dist: aioboto3>=13.0.0
Requires-Dist: alembic>=1.13.0
Requires-Dist: anthropic>=0.40.0
Requires-Dist: asyncpg>=0.29.0
Requires-Dist: claude-agent-sdk>=0.1.18
Requires-Dist: fastapi>=0.109.0
Requires-Dist: google-generativeai>=0.8.0
Requires-Dist: harbor==0.1.45
Requires-Dist: httpx>=0.27.0
Requires-Dist: openai>=1.50.0
Requires-Dist: pgqueuer>=0.18.0
Requires-Dist: pydantic-settings>=2.2.0
Requires-Dist: pydantic>=2.6.0
Requires-Dist: python-jose[cryptography]>=3.3.0
Requires-Dist: python-multipart>=0.0.9
Requires-Dist: rich>=13.7.0
Requires-Dist: sqlalchemy[asyncio]>=2.0.0
Requires-Dist: swegen>=0.1.1
Requires-Dist: typer>=0.12.0
Requires-Dist: uvicorn>=0.27.0
Provides-Extra: all
Requires-Dist: pre-commit>=4.0.0; extra == 'all'
Requires-Dist: pytest-asyncio>=0.23.0; extra == 'all'
Requires-Dist: pytest>=8.0.0; extra == 'all'
Requires-Dist: twine>=6.0.0; extra == 'all'
Provides-Extra: dev
Requires-Dist: pre-commit>=4.0.0; extra == 'dev'
Requires-Dist: pytest-asyncio>=0.23.0; extra == 'dev'
Requires-Dist: pytest>=8.0.0; extra == 'dev'
Requires-Dist: twine>=6.0.0; extra == 'dev'
Description-Content-Type: text/markdown

# Oddish Core Library

This README is focused on implementation details for the `oddish` package.

## Deep Technical Documentation

This covers architecture, configuration, and operational details.

## Architecture

```
┌─────────────────────────────────────────────────────────────┐
│  CLI (oddish run/status) or API Client                      │
│  - Uses env vars for API URL + auth                         │
│  - Submits tasks via HTTP                                   │
│  - Watches tasks or experiments via CLI status              │
└─────────────────────────────────────────────────────────────┘
                              │
                              ▼
┌─────────────────────────────────────────────────────────────┐
│  FastAPI Server (python -m oddish.api)                      │
│  - POST /tasks/upload, /tasks/sweep                         │
│  - GET /tasks, /tasks/{id}, /trials/{id}/logs               │
│  - Auto-starts workers by default                           │
└─────────────────────────────────────────────────────────────┘
                              │
                              ▼
┌─────────────────────────────────────────────────────────────┐
│  Postgres                                                   │
│  - experiments table (grouping + sharing metadata)          │
│  - tasks table (task metadata + verdict)                    │
│  - trials table (runs + analysis)                           │
│  - pgqueuer tables (job queue)                              │
└─────────────────────────────────────────────────────────────┘
                              │
                              ▼
┌─────────────────────────────────────────────────────────────┐
│  PGQueuer Workers                                           │
│  - Poll queue via SELECT FOR UPDATE SKIP LOCKED             │
│  - Queue-key concurrency limits                             │
│  - Execute trials, analyses, and verdict jobs               │
└─────────────────────────────────────────────────────────────┘
```

## Local Development

### Quick start (recommended)

```bash
cp env.example .env
docker compose up -d db
uv sync
uv run python -m oddish.db setup
uv run python -m oddish.api
```

This starts Postgres, runs the API with workers, and is a good baseline for local development.

### CLI configuration

The CLI can talk to either:

- **Local API**: `http://localhost:8000` (self-hosted)
- **Hosted API** (optional): `https://abundant-ai--api.modal.run`

For local use, point the CLI at your server:

```bash
export ODDISH_API_URL="http://localhost:8000"
```

The local API does not enforce auth by default. For the hosted API, set an API key:

```bash
export ODDISH_API_URL="https://abundant-ai--api.modal.run"
export ODDISH_API_KEY="ok_..."
```

### CLI config precedence

The CLI resolves API settings in this order:

1. `ODDISH_API_URL` / `ODDISH_API_KEY` / `ODDISH_DASHBOARD_URL`
2. `ODDISH_DEFAULT_API_URL` / `ODDISH_DEFAULT_DASHBOARD_URL`
3. Built-in defaults:
   - API: `https://abundant-ai--api.modal.run`
   - Dashboard: `https://www.oddish.app`

### Database setup commands

```bash
uv run python -m oddish.db setup            # Full setup (Alembic + PGQueuer)
uv run python -m oddish.db init             # Run Alembic migrations only
uv run python -m oddish.db install-pgqueuer # Install PGQueuer tables only
uv run python -m oddish.db reset            # Drop and recreate all tables
uv run python -m oddish.db purge            # Delete all data (preserves schema)
```

### API flags

```bash
# Set queue-key concurrency
uv run python -m oddish.api --n-concurrent '{"openai/gpt-5.2": 8, "anthropic/claude-sonnet-4-5": 8}'

# Custom host/port
uv run python -m oddish.api --host 0.0.0.0 --port 9000
```

## API Endpoints

| Method | Endpoint | Description |
|--------|----------|-------------|
| GET | `/health` | Health check |
| GET | `/docs` | Swagger UI |
| GET | `/tasks` | List tasks |
| GET | `/tasks/{task_id}` | Task details with trials |
| DELETE | `/tasks/{task_id}` | Delete a task and its trials |
| DELETE | `/experiments/{experiment_id}` | Delete an experiment and its tasks |
| PATCH | `/experiments/{experiment_id}` | Update experiment name |
| POST | `/tasks/upload` | Upload task tarball |
| POST | `/tasks/sweep` | Create evaluation sweep |
| GET | `/tasks/{task_id}/trials/{index}` | Fetch trial by index |
| GET | `/trials/{trial_id}/logs` | Fetch trial logs |
| GET | `/trials/{trial_id}/result` | Fetch `result.json` |

## Docker Compose

The `docker-compose.yml` orchestrates local development:

```bash
# Database only (for local Python dev)
docker compose up -d db

# Full stack (containerized)
docker compose up -d db api worker

# One-time database initialization
docker compose run --rm db-init
```

Services:

| Service | Description |
|---------|-------------|
| `db` | Postgres 16 |
| `api` | FastAPI server (`python -m oddish.api`) |
| `worker` | Standalone worker (`python -m oddish.workers.queue.worker`) |
| `db-init` | One-time setup: runs Alembic migrations + PGQueuer install |

## Configuration

Oddish loads environment variables from `.env` by default (via Pydantic Settings with the `ODDISH_` prefix).

### Database URL

Both formats supported:

```bash
DATABASE_URL=postgresql+asyncpg://oddish:oddish@localhost:5432/oddish
ODDISH_DATABASE_URL=postgresql+asyncpg://...  # Alternative
```

`DATABASE_URL` takes precedence over `ODDISH_DATABASE_URL`.

### Storage

**Local (default):**

- Tasks: `/tmp/oddish-tasks`
- Harbor artifacts: `/tmp/harbor-jobs`

**S3/R2 (production):**

```bash
ODDISH_S3_ENABLED=true
ODDISH_S3_BUCKET=data
ODDISH_S3_ACCESS_KEY=...
ODDISH_S3_SECRET_KEY=...
ODDISH_S3_ENDPOINT_URL=https://...
```

Task uploads land under `tasks/<task_id>/`. Trial artifacts are uploaded under
`tasks/<task_id>/trials/<trial_id>/` when possible, with a fallback of
`trials/<trial_id>/` for legacy IDs.

### Execution Environments

Oddish runs Harbor tasks in a sandboxed environment.

CLI behavior when `--env` is omitted:

- Local API URL (`localhost`) defaults to `docker`
- Hosted Modal API URL (`*.modal.run`) defaults to `modal`
- Other remote API URLs default to `docker`

You can always override per task with:
`oddish run --env {docker|daytona|e2b|modal|runloop|gke}`.

### Queue-Key Routing

Oddish routes jobs by **queue key** (normalized model string) for PGQueuer entrypoints.
Agent names still map to provider buckets for compatibility/attribution, but queueing
uses `get_queue_key_for_trial(agent, model)` and defaults to the agent fallback only
when no model is provided.

### Concurrency Control

Queue-key concurrency is fixed at API startup (not per job).

Order of precedence:

1. **Manual API startup:** `python -m oddish.api --n-concurrent '{"openai/gpt-5.2": 8}'`
2. **Default:** `ODDISH_DEFAULT_MODEL_CONCURRENCY` (with optional model overrides)

For self-hosted setups, set concurrency on API startup:
```bash
uv run python -m oddish.api --n-concurrent '{"openai/gpt-5.2": 8, "anthropic/claude-sonnet-4-5": 8}'
```

Changing concurrency requires restarting the API process.

### LLM API Keys

Only set keys for providers you use:

```bash
ANTHROPIC_API_KEY=sk-...
OPENAI_API_KEY=sk-...
GEMINI_API_KEY=...
```

### Sandbox Provider Keys

Set keys for the sandbox environments you use:

```bash
DAYTONA_API_KEY=...
MODAL_TOKEN_ID=...
MODAL_TOKEN_SECRET=...
```

## Execution Pipeline

Tasks move through a multi-stage pipeline when `run_analysis` is enabled:

1. **Trials** run for each agent/model pair (status: pending → queued → running → success/failed).
2. **Analyses** run per trial after completion to classify outcomes.
3. **Verdict** runs once per task to summarize analyses.

Each stage is a PGQueuer job routed through queue-key entrypoints. Task status
progresses from `pending` → `running` → `analyzing` → `verdict_pending` → `completed`
(or `failed` on terminal error).

## Experiments

Every task belongs to an experiment. If no experiment is provided, Oddish
generates a short, human-friendly name (`oddish.experiment.generate_experiment_name`).
The CLI can watch an experiment with `oddish status --experiment <id>`.

## Workers

### Auto-start behavior

By default, `python -m oddish.api` spawns workers in background threads.

For separate worker processes (for scaling), run:

```bash
uv run python -m oddish.workers.queue.worker
```

In Docker Compose, this maps to the dedicated `worker` service.

### How PGQueuer works

Workers claim jobs atomically via Postgres:

```sql
SELECT * FROM pgqueuer
WHERE status = 'queued' AND entrypoint = 'openai/gpt-5.2'
ORDER BY priority DESC
LIMIT 1
FOR UPDATE SKIP LOCKED;

UPDATE pgqueuer SET status = 'processing' WHERE id = ?;
```

- `FOR UPDATE`: Locks the row
- `SKIP LOCKED`: Other workers skip locked rows
- **Result:** Each job claimed by exactly one worker

### Concurrency enforcement

PGQueuer checks processing count before claiming:

```sql
SELECT COUNT(*) FROM pgqueuer
WHERE entrypoint = 'openai/gpt-5.2' AND status = 'processing';
```

If count >= limit (e.g., 8), worker waits. Concurrency is database state, not worker count.

### Job routing

Queue entrypoints are created per queue key (typically model identifiers).
Each entrypoint handles jobs with `job_type` of `trial`, `analysis`, or `verdict`.

## CLI Reference

```bash
# Point at local API
export ODDISH_API_URL="http://localhost:8000"

# Run a task
oddish run ./my-task -a claude-code -m claude-sonnet-4-5

# Run a sweep
oddish run -d terminalbench@2.0 -c sweep.yaml

# Optional run flags
oddish run ./my-task --run-analysis --env daytona --override-cpus 4

# Pass environment variables and kwargs to the agent
oddish run ./my-task -a claude-code -m claude-sonnet-4-5 \
  --ae AWS_REGION=us-east-1 --ak max_thinking_tokens=8000

# Force rebuild and collect artifacts
oddish run ./my-task -a claude-code --force-build --artifact /workspace/output.txt

# Monitor
oddish status
oddish status <task_id> --watch
oddish status --experiment <experiment_id> --watch

# Cleanup
oddish clean <task_id>
oddish clean --experiment <id>
oddish clean --all-experiments
```

### Sweep Config Files

A sweep config file (YAML or JSON, passed via `oddish run -c sweep.yaml`) defines
which agents and models to evaluate:

```yaml
agents:
  - name: claude-code
    model_name: anthropic/claude-sonnet-4-5
    n_trials: 3
    env:                        # optional: agent env vars
      CUSTOM_VAR: "value"
    kwargs:                     # optional: agent kwargs
      max_thinking_tokens: 8000

  - name: codex
    model_name: openai/gpt-5.2
    n_trials: 3
    timeout_minutes: 120        # optional: per-agent timeout

# Task source (pick one):
path: ./my-task                 # local task or dataset directory
dataset: swebench@1.0           # registry dataset

# Optional filtering:
task_names: ["task-*"]
exclude_task_names: ["*-slow"]
n_tasks: 10

# Optional fields:
environment: daytona
priority: low
experiment_id: exp_123
```

### Harbor Execution Config

All Harbor execution settings (environment resources, verifier config, artifacts)
are passed via a nested `harbor` field in the API. This directly uses Harbor's
native `EnvironmentConfig`, `VerifierConfig`, and `ArtifactConfig` types:

```json
{
  "task_id": "abc123",
  "configs": [{"agent": "claude-code", "model": "claude-sonnet-4-5", "n_trials": 3}],
  "user": "alice",
  "harbor": {
    "environment": {
      "override_cpus": 4,
      "override_memory_mb": 8192,
      "override_gpus": 1,
      "kwargs": {
        "network_block_all": false,
        "sandbox_timeout_secs": 86400
      }
    },
    "verifier": {
      "disable": true
    },
    "artifacts": ["/workspace/output.txt"],
    "docker_image": "my-registry/my-image:latest"
  }
}
```

Per-trial agent overrides (env vars, kwargs, timeouts) use `agent_config`:

```json
{
  "configs": [
    {
      "agent": "claude-code",
      "model": "claude-sonnet-4-5",
      "n_trials": 3,
      "agent_config": {
        "env": {"CUSTOM_VAR": "value"},
        "kwargs": {"max_thinking_tokens": 8000},
        "override_timeout_sec": 7200
      }
    }
  ]
}
```

### GitHub Actions Integration

The CLI supports JSON output for CI pipelines:

```bash
oddish run ./tasks/* -a codex --n-trials 1 --json

# Output:
# {
#   "experiment": "random-words-123",
#   "experiment_url": "...",
#   "total_trials": 3,
#   "tasks": [
#     {"id": "task-abc123", "trials_count": 1, "url": "..."},
#     ...
#   ]
# }
```

Environment variables for CI:
- `ODDISH_API_URL`: API endpoint (your self-hosted URL, or the hosted API)
- `ODDISH_API_KEY`: API token (required for the hosted API)

## Repository Structure

```
oddish/
├── src/oddish/
│   ├── __init__.py          # Public API (lazy-loaded exports)
│   ├── config.py            # Settings, provider mapping
│   ├── schemas.py           # Pydantic request/response models (HarborConfig, TaskSubmission, etc.)
│   ├── queue.py             # Task creation, queue orchestration
│   ├── experiment.py        # Experiment name generation
│   ├── infra.py             # Docker/infrastructure helpers
│   ├── api/
│   │   ├── __init__.py      # FastAPI app + endpoint wiring
│   │   ├── endpoints.py     # Core endpoint logic
│   │   ├── helpers.py       # Response builders
│   │   ├── tasks.py         # Task upload handling
│   │   └── trial_io.py      # Trial logs/result reading
│   ├── cli/
│   │   ├── __init__.py      # Typer app entry point
│   │   ├── run.py           # Run command (task submission)
│   │   ├── status.py        # Status command (monitoring)
│   │   ├── clean.py         # Clean command (deletion)
│   │   ├── api.py           # HTTP client helpers
│   │   ├── config.py        # API URL/auth resolution
│   │   └── infra.py         # Local infrastructure helpers
│   ├── db/
│   │   ├── __init__.py      # DB exports
│   │   ├── __main__.py      # CLI: python -m oddish.db
│   │   ├── models.py        # SQLAlchemy models (Experiment, Task, Trial)
│   │   ├── connection.py    # Engine, session factory, pool management
│   │   └── storage.py       # S3/local storage client
│   └── workers/
│       ├── harbor_runner.py # Harbor task executor + artifact upload
│       └── queue/
│           ├── queue_manager.py    # PGQueuer setup + entrypoints
│           ├── worker.py           # Standalone worker entry point
│           ├── trial_handler.py    # Trial execution handler
│           ├── analysis_handler.py # Post-trial analysis handler
│           ├── verdict_handler.py  # Task-level verdict handler
│           ├── db_helpers.py       # Worker DB utilities
│           └── shared.py           # Shared worker utilities
│
├── alembic/                 # Database migrations
├── alembic.ini              # Alembic configuration
├── docker-compose.yml       # Local dev orchestration
├── env.example              # Example .env file
├── pyproject.toml           # Package config and dependencies
└── README.md
```

## Using as a Library

Oddish can be imported as a library in your own services:

```python
# Database models and sessions
from oddish.db import TaskModel, TrialModel, get_session, init_db

# Queue operations
from oddish.queue import create_task, get_task_with_trials, get_queue_stats

# Worker logic
from oddish.workers.queue import create_queue_manager

# Configuration
from oddish.config import settings

# Schemas
from oddish.schemas import TaskSubmission, TaskSweepSubmission, TrialSpec, HarborConfig
```

## Database Migrations

Oddish uses Alembic for schema management. The version table is `alembic_version_oddish` (to avoid conflicts if you run your own Alembic migrations in the same database).

PGQueuer tables are managed separately via `oddish.db install-pgqueuer`.

```bash
# Run all migrations
uv run alembic upgrade head

# Check current version
uv run alembic current

# Full setup (migrations + PGQueuer)
uv run python -m oddish.db setup
```

## Troubleshooting

### Port conflicts

| Service | Port |
|---------|------|
| Postgres | 5432 |
| API | 8000 |

If ports are in use, stop the conflicting process or change the port.

### Database connection errors

```bash
# Verify Postgres is running
docker compose ps

# Test connection
psql $DATABASE_URL -c "SELECT 1"

# Check migrations
uv run alembic current
uv run alembic upgrade head
```

### Tasks stuck in "queued"

1. Check workers are running and API is healthy:
   ```bash
   curl http://localhost:8000/health
   oddish status
   ```

2. Check queue-key concurrency limits (set at API startup) and worker logs

3. Check for errors in API logs

### Harbor execution failures

1. Verify the sandbox environment is available (Docker running, Daytona key set, etc.)
2. Check LLM API key is set for the provider
3. Check trial error message:
   ```bash
   curl http://localhost:8000/tasks/<task_id> | jq '.trials[].error_message'
   ```
