weco 0.2.12__tar.gz → 0.2.14__tar.gz

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.
Files changed (47) hide show
  1. weco-0.2.14/.github/workflows/release.yml +151 -0
  2. {weco-0.2.12 → weco-0.2.14}/.gitignore +4 -1
  3. weco-0.2.14/.repomixignore +4 -0
  4. {weco-0.2.12 → weco-0.2.14}/PKG-INFO +61 -23
  5. {weco-0.2.12 → weco-0.2.14}/README.md +60 -22
  6. {weco-0.2.12 → weco-0.2.14}/pyproject.toml +1 -1
  7. weco-0.2.14/weco/__init__.py +14 -0
  8. {weco-0.2.12 → weco-0.2.14}/weco/api.py +19 -6
  9. weco-0.2.14/weco/auth.py +64 -0
  10. weco-0.2.14/weco/cli.py +583 -0
  11. {weco-0.2.12 → weco-0.2.14}/weco/panels.py +20 -2
  12. {weco-0.2.12 → weco-0.2.14}/weco.egg-info/PKG-INFO +61 -23
  13. {weco-0.2.12 → weco-0.2.14}/weco.egg-info/SOURCES.txt +2 -0
  14. weco-0.2.12/.github/workflows/release.yml +0 -106
  15. weco-0.2.12/weco/__init__.py +0 -4
  16. weco-0.2.12/weco/cli.py +0 -378
  17. {weco-0.2.12 → weco-0.2.14}/.github/workflows/lint.yml +0 -0
  18. {weco-0.2.12 → weco-0.2.14}/LICENSE +0 -0
  19. {weco-0.2.12 → weco-0.2.14}/examples/cuda/README.md +0 -0
  20. {weco-0.2.12 → weco-0.2.14}/examples/cuda/evaluate.py +0 -0
  21. {weco-0.2.12 → weco-0.2.14}/examples/cuda/guide.md +0 -0
  22. {weco-0.2.12 → weco-0.2.14}/examples/cuda/optimize.py +0 -0
  23. {weco-0.2.12 → weco-0.2.14}/examples/hello-kernel-world/evaluate.py +0 -0
  24. {weco-0.2.12 → weco-0.2.14}/examples/hello-kernel-world/optimize.py +0 -0
  25. {weco-0.2.12 → weco-0.2.14}/examples/metal/README.md +0 -0
  26. {weco-0.2.12 → weco-0.2.14}/examples/metal/evaluate.py +0 -0
  27. {weco-0.2.12 → weco-0.2.14}/examples/metal/examples.rst +0 -0
  28. {weco-0.2.12 → weco-0.2.14}/examples/metal/optimize.py +0 -0
  29. {weco-0.2.12 → weco-0.2.14}/examples/prompt/README.md +0 -0
  30. {weco-0.2.12 → weco-0.2.14}/examples/prompt/eval.py +0 -0
  31. {weco-0.2.12 → weco-0.2.14}/examples/prompt/optimize.py +0 -0
  32. {weco-0.2.12 → weco-0.2.14}/examples/prompt/prompt_guide.md +0 -0
  33. {weco-0.2.12 → weco-0.2.14}/examples/spaceship-titanic/README.md +0 -0
  34. {weco-0.2.12 → weco-0.2.14}/examples/spaceship-titanic/baseline.py +0 -0
  35. {weco-0.2.12 → weco-0.2.14}/examples/spaceship-titanic/evaluate.py +0 -0
  36. {weco-0.2.12 → weco-0.2.14}/examples/spaceship-titanic/optimize.py +0 -0
  37. {weco-0.2.12 → weco-0.2.14}/examples/spaceship-titanic/requirements-test.txt +0 -0
  38. {weco-0.2.12 → weco-0.2.14}/examples/spaceship-titanic/utils.py +0 -0
  39. {weco-0.2.12 → weco-0.2.14}/examples/triton/README.md +0 -0
  40. {weco-0.2.12 → weco-0.2.14}/examples/triton/evaluate.py +0 -0
  41. {weco-0.2.12 → weco-0.2.14}/examples/triton/optimize.py +0 -0
  42. {weco-0.2.12 → weco-0.2.14}/setup.cfg +0 -0
  43. {weco-0.2.12 → weco-0.2.14}/weco/utils.py +0 -0
  44. {weco-0.2.12 → weco-0.2.14}/weco.egg-info/dependency_links.txt +0 -0
  45. {weco-0.2.12 → weco-0.2.14}/weco.egg-info/entry_points.txt +0 -0
  46. {weco-0.2.12 → weco-0.2.14}/weco.egg-info/requires.txt +0 -0
  47. {weco-0.2.12 → weco-0.2.14}/weco.egg-info/top_level.txt +0 -0
weco-0.2.14/.github/workflows/release.yml ADDED
@@ -0,0 +1,151 @@
1
+ name: Release
2
+
3
+ on:
4
+ # Fires when the *Lint and Format Code* workflow that was
5
+ # started by a push to main has finished (any conclusion)
6
+ workflow_run:
7
+ workflows: ["Lint and Format Code"]
8
+ branches: [main]
9
+ types: [completed]
10
+
11
+ jobs:
12
+ # ────────────────────────────────────────────────────────────────────
13
+ # 1) Pre-check — decide whether we really need to release
14
+ # ────────────────────────────────────────────────────────────────────
15
+ pre-check:
16
+ name: Detect new version
17
+ if: ${{ github.event.workflow_run.conclusion == 'success' }} # gate #1
18
+ runs-on: ubuntu-latest
19
+ outputs:
20
+ release_needed: ${{ steps.version_diff.outputs.release_needed }}
21
+ version: ${{ steps.version_diff.outputs.version }}
22
+
23
+ steps:
24
+ - name: Checkout code
25
+ uses: actions/checkout@v4
26
+ with:
27
+ ref: ${{ github.event.workflow_run.head_sha }}
28
+ fetch-depth: 0 # we need history to read the previous commit
29
+
30
+ - name: Compare versions
31
+ id: version_diff
32
+ shell: bash
33
+ # gate #2 happens here → sets release_needed=true/false
34
+ run: |
35
+ # version in current pyproject.toml
36
+ NEW_VERSION=$(grep -Po '(?<=^version = ")[^"]+' pyproject.toml)
37
+ # version in the previous commit's pyproject.toml (if the file existed)
38
+ BASE_COMMIT=$(git rev-parse "$GITHUB_SHA"^)
39
+ if git cat-file -e "$BASE_COMMIT":pyproject.toml 2>/dev/null; then
40
+ OLD_VERSION=$(git show "$BASE_COMMIT":pyproject.toml \
41
+ | grep -Po '(?<=^version = ")[^"]+' || true)
42
+ else
43
+ OLD_VERSION=""
44
+ fi
45
+
46
+ echo "Previous version: $OLD_VERSION"
47
+ echo "Current version: $NEW_VERSION"
48
+
49
+ if [[ "$NEW_VERSION" != "$OLD_VERSION" ]]; then
50
+ echo "release_needed=true" >>"$GITHUB_OUTPUT"
51
+ else
52
+ echo "release_needed=false" >>"$GITHUB_OUTPUT"
53
+ fi
54
+ echo "version=$NEW_VERSION" >>"$GITHUB_OUTPUT"
55
+
56
+ # ────────────────────────────────────────────────────────────────────
57
+ # 2) Build
58
+ # ────────────────────────────────────────────────────────────────────
59
+ build:
60
+ name: Build distribution 📦
61
+ needs: pre-check
62
+ if: ${{ needs.pre-check.outputs.release_needed == 'true' }} # gate #3
63
+ runs-on: ubuntu-latest
64
+
65
+ steps:
66
+ - name: Checkout code
67
+ uses: actions/checkout@v4
68
+ with:
69
+ ref: ${{ github.event.workflow_run.head_sha }}
70
+
71
+ - name: Set up Python
72
+ uses: actions/setup-python@v5
73
+ with:
74
+ python-version: "3.12"
75
+
76
+ - name: Install build dependencies
77
+ run: python3 -m pip install --user build
78
+
79
+ - name: Build sdist & wheel
80
+ run: python3 -m build
81
+
82
+ - name: Store the distribution packages
83
+ uses: actions/upload-artifact@v4
84
+ with:
85
+ name: python-package-distributions
86
+ path: dist/
87
+
88
+ # ────────────────────────────────────────────────────────────────────
89
+ # 3) Publish to PyPI
90
+ # ────────────────────────────────────────────────────────────────────
91
+ publish-to-pypi:
92
+ name: Publish Python 🐍 distribution to PyPI
93
+ needs: [pre-check, build]
94
+ if: ${{ needs.pre-check.outputs.release_needed == 'true' }}
95
+ runs-on: ubuntu-latest
96
+ environment:
97
+ name: release
98
+ url: https://pypi.org/p/weco
99
+ permissions:
100
+ id-token: write
101
+
102
+ steps:
103
+ - name: Download the dists
104
+ uses: actions/download-artifact@v4
105
+ with:
106
+ name: python-package-distributions
107
+ path: dist/
108
+
109
+ - name: Publish to PyPI
110
+ uses: pypa/gh-action-pypi-publish@release/v1
111
+
112
+ # ────────────────────────────────────────────────────────────────────
113
+ # 4) GitHub Release
114
+ # ────────────────────────────────────────────────────────────────────
115
+ github-release:
116
+ name: Create GitHub Release
117
+ needs: [pre-check, publish-to-pypi]
118
+ if: ${{ needs.pre-check.outputs.release_needed == 'true' }}
119
+ runs-on: ubuntu-latest
120
+ permissions:
121
+ contents: write
122
+ id-token: write
123
+
124
+ steps:
125
+ - name: Download the dists
126
+ uses: actions/download-artifact@v4
127
+ with:
128
+ name: python-package-distributions
129
+ path: dist/
130
+
131
+ - name: Sign dists with Sigstore
132
+ uses: sigstore/gh-action-sigstore-python@v3.0.0
133
+ with:
134
+ inputs: |
135
+ ./dist/*.tar.gz
136
+ ./dist/*.whl
137
+
138
+ - name: Create GitHub Release
139
+ env:
140
+ GITHUB_TOKEN: ${{ github.token }}
141
+ run: |
142
+ gh release create "v${{ needs.pre-check.outputs.version }}" \
143
+ --repo "${{ github.repository }}" \
144
+ --notes ""
145
+
146
+ - name: Upload artefacts to Release
147
+ env:
148
+ GITHUB_TOKEN: ${{ github.token }}
149
+ run: |
150
+ gh release upload "v${{ needs.pre-check.outputs.version }}" dist/** \
151
+ --repo "${{ github.repository }}"
{weco-0.2.12 → weco-0.2.14}/.gitignore RENAMED
@@ -70,4 +70,7 @@ etc/
70
70
  digest.txt
71
71
  .runs/
72
72
 
73
- *.pyc
73
+ *.pyc
74
+
75
+ # Repomix
76
+ repomix-output.*
weco-0.2.14/.repomixignore ADDED
@@ -0,0 +1,4 @@
1
+ examples/
2
+ LICENSE
3
+ .gitignore
4
+ .repomixignore
{weco-0.2.12 → weco-0.2.14}/PKG-INFO RENAMED
@@ -1,6 +1,6 @@
1
1
  Metadata-Version: 2.4
2
2
  Name: weco
3
- Version: 0.2.12
3
+ Version: 0.2.14
4
4
  Summary: Documentation for `weco`, a CLI for using Weco AI's code optimizer.
5
5
  Author-email: Weco AI Team <contact@weco.ai>
6
6
  License: MIT
@@ -32,7 +32,7 @@ Example applications include:
32
32
 
33
33
  - **GPU Kernel Optimization**: Reimplement PyTorch functions using CUDA, Triton or Metal, optimizing for `latency`, `throughput`, or `memory_bandwidth`.
34
34
  - **Model Development**: Tune feature transformations or architectures, optimizing for `validation_accuracy`, `AUC`, or `Sharpe Ratio`.
35
- - **Prompt Engineering**: Refine prompts for LLMs, optimizing for `win_rate`, `relevance`, or `format_adherence`
35
+ - **Prompt Engineering**: Refine prompts for LLMs, optimizing for `win_rate`, `relevance`, or `format_adherence`
36
36
 
37
37
  https://github.com/user-attachments/assets/cb724ef1-bff6-4757-b457-d3b2201ede81
38
38
 
@@ -42,7 +42,7 @@ https://github.com/user-attachments/assets/cb724ef1-bff6-4757-b457-d3b2201ede81
42
42
 
43
43
  The `weco` CLI leverages a tree search approach guided by Large Language Models (LLMs) to iteratively explore and refine your code. It automatically applies changes, runs your evaluation script, parses the results, and proposes further improvements based on the specified goal.
44
44
 
45
- ![image](https://github.com/user-attachments/assets/a6ed63fa-9c40-498e-aa98-a873e5786509)
45
+ [image](https://github.com/user-attachments/assets/a6ed63fa-9c40-498e-aa98-a873e5786509)
46
46
 
47
47
  ---
48
48
 
@@ -54,17 +54,38 @@ The `weco` CLI leverages a tree search approach guided by Large Language Models
54
54
  pip install weco
55
55
  ```
56
56
 
57
- 2. **Configure API Keys:**
57
+ 2. **Set Up LLM API Keys (Required):**
58
58
 
59
- Set the appropriate environment variables for your desired language model provider:
59
+ `weco` requires API keys for the Large Language Models (LLMs) it uses internally. You **must** provide these keys via environment variables:
60
60
 
61
- - **OpenAI:** `export OPENAI_API_KEY="your_key_here"`
62
- - **Anthropic:** `export ANTHROPIC_API_KEY="your_key_here"`
63
- - **Google DeepMind:** `export GEMINI_API_KEY="your_key_here"` (Google AI Studio has a free API usage quota. Create a key [here](https://aistudio.google.com/apikey) to use weco for free.)
61
+ - **OpenAI:** `export OPENAI_API_KEY="your_key_here"`
62
+ - **Anthropic:** `export ANTHROPIC_API_KEY="your_key_here"`
63
+ - **Google DeepMind:** `export GEMINI_API_KEY="your_key_here"` (Google AI Studio has a free API usage quota. Create a key [here](https://aistudio.google.com/apikey) to use `weco` for free.)
64
+
65
+ The optimization process will fail if the necessary keys for the chosen model are not found in your environment.
66
+
67
+ 3. **Log In to Weco (Optional):**
68
+
69
+ To associate your optimization runs with your Weco account and view them on the Weco dashboard, you can log in. `weco` uses a device authentication flow:
70
+
71
+ - When you first run `weco run`, you'll be prompted if you want to log in or proceed anonymously.
72
+ - If you choose to log in (by pressing `l`), you'll be shown a URL and `weco` will attempt to open it in your default web browser.
73
+ - You then authenticate in the browser. Once authenticated, the CLI will detect this and complete the login.
74
+ - This saves a Weco-specific API key locally (typically at `~/.config/weco/credentials.json`).
75
+
76
+ If you choose to skip login (by pressing Enter or `s`), `weco` will still function using the environment variable LLM keys, but the run history will not be linked to a Weco account.
77
+
78
+ To log out and remove your saved Weco API key, use the `weco logout` command.
64
79
 
65
80
  ---
66
81
 
67
82
  ## Usage
83
+
84
+ The CLI has two main commands:
85
+
86
+ - `weco run`: Initiates the code optimization process.
87
+ - `weco logout`: Logs you out of your Weco account.
88
+
68
89
  <div style="background-color: #fff3cd; border: 1px solid #ffeeba; padding: 15px; border-radius: 4px; margin-bottom: 15px;">
69
90
  <strong>⚠️ Warning: Code Modification</strong><br>
70
91
  <code>weco</code> directly modifies the file specified by <code>--source</code> during the optimization process. It is <strong>strongly recommended</strong> to use version control (like Git) to track changes and revert if needed. Alternatively, ensure you have a backup of your original file before running the command. Upon completion, the file will contain the best-performing version of the code found during the run.
@@ -72,7 +93,11 @@ The `weco` CLI leverages a tree search approach guided by Large Language Models
72
93
 
73
94
  ---
74
95
 
75
- ### Example: Optimizing Simple PyTorch Operations
96
+ ### `weco run` Command
97
+
98
+ This command starts the optimization process.
99
+
100
+ **Example: Optimizing Simple PyTorch Operations**
76
101
 
77
102
  This basic example shows how to optimize a simple PyTorch function for speedup.
78
103
 
@@ -86,7 +111,7 @@ cd examples/hello-kernel-world
86
111
  pip install torch
87
112
 
88
113
  # Run Weco
89
- weco --source optimize.py \
114
+ weco run --source optimize.py \
90
115
  --eval-command "python evaluate.py --solution-path optimize.py --device cpu" \
91
116
  --metric speedup \
92
117
  --maximize true \
@@ -99,19 +124,29 @@ weco --source optimize.py \
99
124
 
100
125
  ---
101
126
 
102
- ### Command Line Arguments
127
+ **Arguments for `weco run`:**
103
128
 
104
- | Argument | Description | Required |
105
- | :-------------------------- | :----------------------------------------------------------------------------------------------------------------------------------------------------------------------- | :------- |
106
- | `--source` | Path to the source code file that will be optimized (e.g., `optimize.py`). | Yes |
107
- | `--eval-command` | Command to run for evaluating the code in `--source`. This command should print the target `--metric` and its value to the terminal (stdout/stderr). See note below. | Yes |
108
- | `--metric` | The name of the metric you want to optimize (e.g., 'accuracy', 'speedup', 'loss'). This metric name should match what's printed by your `--eval-command`. | Yes |
109
- | `--maximize` | Whether to maximize (`true`) or minimize (`false`) the metric. | Yes |
110
- | `--steps` | Number of optimization steps (LLM iterations) to run. | Yes |
111
- | `--model` | Model identifier for the LLM to use (e.g., `gpt-4o`, `claude-3.7-sonnet`). Recommended models to try include `o4-mini`, and `gemini-2.5-pro-exp-03-25`.| Yes |
112
- | `--additional-instructions` | (Optional) Natural language description of specific instructions OR path to a file containing detailed instructions to guide the LLM. | No |
113
- | `--log-dir` | (Optional) Path to the directory to log intermediate steps and final optimization result. Defaults to `.runs/`. | No |
114
- | `--preserve-source` | (Optional) If set, do not overwrite the original `--source` file. Modifications and the best solution will still be saved in the `--log-dir`. | No |
129
+ | Argument | Description | Required |
130
+ | :-------------------------- | :------------------------------------------------------------------------------------------------------------------------------------------------------------------------ | :------- |
131
+ | `--source` | Path to the source code file that will be optimized (e.g., `optimize.py`). | Yes |
132
+ | `--eval-command` | Command to run for evaluating the code in `--source`. This command should print the target `--metric` and its value to the terminal (stdout/stderr). See note below. | Yes |
133
+ | `--metric` | The name of the metric you want to optimize (e.g., 'accuracy', 'speedup', 'loss'). This metric name should match what's printed by your `--eval-command`. | Yes |
134
+ | `--maximize` | Whether to maximize (`true`) or minimize (`false`) the metric. | Yes |
135
+ | `--steps` | Number of optimization steps (LLM iterations) to run. | Yes |
136
+ | `--model` | Model identifier for the LLM to use (e.g., `gpt-4o`, `claude-3.5-sonnet`). Recommended models to try include `o3-mini`, `claude-3-haiku`, and `gemini-2.5-pro-exp-03-25`. | Yes |
137
+ | `--additional-instructions` | (Optional) Natural language description of specific instructions OR path to a file containing detailed instructions to guide the LLM. | No |
138
+ | `--log-dir` | (Optional) Path to the directory to log intermediate steps and final optimization result. Defaults to `.runs/`. | No |
139
+ | `--preserve-source` | (Optional) If set, do not overwrite the original `--source` file. Modifications and the best solution will still be saved in the `--log-dir`. | No |
140
+
141
+ ---
142
+
143
+ ### `weco logout` Command
144
+
145
+ This command logs you out by removing the locally stored Weco API key.
146
+
147
+ ```bash
148
+ weco logout
149
+ ```
115
150
 
116
151
  ---
117
152
 
@@ -120,6 +155,7 @@ weco --source optimize.py \
120
155
  Weco, powered by the AIDE algorithm, optimizes code iteratively based on your evaluation results. Achieving significant improvements, especially on complex research-level tasks, often requires substantial exploration time.
121
156
 
122
157
  The following plot from the independent [Research Engineering Benchmark (RE-Bench)](https://metr.org/AI_R_D_Evaluation_Report.pdf) report shows the performance of AIDE (the algorithm behind Weco) on challenging ML research engineering tasks over different time budgets.
158
+
123
159
  <p align="center">
124
160
  <img src="https://github.com/user-attachments/assets/ff0e471d-2f50-4e2d-b718-874862f533df" alt="RE-Bench Performance Across Time" width="60%"/>
125
161
  </p>
@@ -146,23 +182,25 @@ Final speedup value = 1.5
146
182
 
147
183
  Weco will parse this output to extract the numerical value (1.5 in this case) associated with the metric name ('speedup').
148
184
 
149
-
150
185
  ## Contributing
151
186
 
152
187
  We welcome contributions! To get started:
153
188
 
154
189
  1. **Fork and Clone the Repository:**
190
+
155
191
  ```bash
156
192
  git clone https://github.com/WecoAI/weco-cli.git
157
193
  cd weco-cli
158
194
  ```
159
195
 
160
196
  2. **Install Development Dependencies:**
197
+
161
198
  ```bash
162
199
  pip install -e ".[dev]"
163
200
  ```
164
201
 
165
202
  3. **Create a Feature Branch:**
203
+
166
204
  ```bash
167
205
  git checkout -b feature/your-feature-name
168
206
  ```
{weco-0.2.12 → weco-0.2.14}/README.md RENAMED
@@ -10,7 +10,7 @@ Example applications include:
10
10
 
11
11
  - **GPU Kernel Optimization**: Reimplement PyTorch functions using CUDA, Triton or Metal, optimizing for `latency`, `throughput`, or `memory_bandwidth`.
12
12
  - **Model Development**: Tune feature transformations or architectures, optimizing for `validation_accuracy`, `AUC`, or `Sharpe Ratio`.
13
- - **Prompt Engineering**: Refine prompts for LLMs, optimizing for `win_rate`, `relevance`, or `format_adherence`
13
+ - **Prompt Engineering**: Refine prompts for LLMs, optimizing for `win_rate`, `relevance`, or `format_adherence`
14
14
 
15
15
  https://github.com/user-attachments/assets/cb724ef1-bff6-4757-b457-d3b2201ede81
16
16
 
@@ -20,7 +20,7 @@ https://github.com/user-attachments/assets/cb724ef1-bff6-4757-b457-d3b2201ede81
20
20
 
21
21
  The `weco` CLI leverages a tree search approach guided by Large Language Models (LLMs) to iteratively explore and refine your code. It automatically applies changes, runs your evaluation script, parses the results, and proposes further improvements based on the specified goal.
22
22
 
23
- ![image](https://github.com/user-attachments/assets/a6ed63fa-9c40-498e-aa98-a873e5786509)
23
+ [image](https://github.com/user-attachments/assets/a6ed63fa-9c40-498e-aa98-a873e5786509)
24
24
 
25
25
  ---
26
26
 
@@ -32,17 +32,38 @@ The `weco` CLI leverages a tree search approach guided by Large Language Models
32
32
  pip install weco
33
33
  ```
34
34
 
35
- 2. **Configure API Keys:**
35
+ 2. **Set Up LLM API Keys (Required):**
36
36
 
37
- Set the appropriate environment variables for your desired language model provider:
37
+ `weco` requires API keys for the Large Language Models (LLMs) it uses internally. You **must** provide these keys via environment variables:
38
38
 
39
- - **OpenAI:** `export OPENAI_API_KEY="your_key_here"`
40
- - **Anthropic:** `export ANTHROPIC_API_KEY="your_key_here"`
41
- - **Google DeepMind:** `export GEMINI_API_KEY="your_key_here"` (Google AI Studio has a free API usage quota. Create a key [here](https://aistudio.google.com/apikey) to use weco for free.)
39
+ - **OpenAI:** `export OPENAI_API_KEY="your_key_here"`
40
+ - **Anthropic:** `export ANTHROPIC_API_KEY="your_key_here"`
41
+ - **Google DeepMind:** `export GEMINI_API_KEY="your_key_here"` (Google AI Studio has a free API usage quota. Create a key [here](https://aistudio.google.com/apikey) to use `weco` for free.)
42
+
43
+ The optimization process will fail if the necessary keys for the chosen model are not found in your environment.
44
+
45
+ 3. **Log In to Weco (Optional):**
46
+
47
+ To associate your optimization runs with your Weco account and view them on the Weco dashboard, you can log in. `weco` uses a device authentication flow:
48
+
49
+ - When you first run `weco run`, you'll be prompted if you want to log in or proceed anonymously.
50
+ - If you choose to log in (by pressing `l`), you'll be shown a URL and `weco` will attempt to open it in your default web browser.
51
+ - You then authenticate in the browser. Once authenticated, the CLI will detect this and complete the login.
52
+ - This saves a Weco-specific API key locally (typically at `~/.config/weco/credentials.json`).
53
+
54
+ If you choose to skip login (by pressing Enter or `s`), `weco` will still function using the environment variable LLM keys, but the run history will not be linked to a Weco account.
55
+
56
+ To log out and remove your saved Weco API key, use the `weco logout` command.
42
57
 
43
58
  ---
44
59
 
45
60
  ## Usage
61
+
62
+ The CLI has two main commands:
63
+
64
+ - `weco run`: Initiates the code optimization process.
65
+ - `weco logout`: Logs you out of your Weco account.
66
+
46
67
  <div style="background-color: #fff3cd; border: 1px solid #ffeeba; padding: 15px; border-radius: 4px; margin-bottom: 15px;">
47
68
  <strong>⚠️ Warning: Code Modification</strong><br>
48
69
  <code>weco</code> directly modifies the file specified by <code>--source</code> during the optimization process. It is <strong>strongly recommended</strong> to use version control (like Git) to track changes and revert if needed. Alternatively, ensure you have a backup of your original file before running the command. Upon completion, the file will contain the best-performing version of the code found during the run.
@@ -50,7 +71,11 @@ The `weco` CLI leverages a tree search approach guided by Large Language Models
50
71
 
51
72
  ---
52
73
 
53
- ### Example: Optimizing Simple PyTorch Operations
74
+ ### `weco run` Command
75
+
76
+ This command starts the optimization process.
77
+
78
+ **Example: Optimizing Simple PyTorch Operations**
54
79
 
55
80
  This basic example shows how to optimize a simple PyTorch function for speedup.
56
81
 
@@ -64,7 +89,7 @@ cd examples/hello-kernel-world
64
89
  pip install torch
65
90
 
66
91
  # Run Weco
67
- weco --source optimize.py \
92
+ weco run --source optimize.py \
68
93
  --eval-command "python evaluate.py --solution-path optimize.py --device cpu" \
69
94
  --metric speedup \
70
95
  --maximize true \
@@ -77,19 +102,29 @@ weco --source optimize.py \
77
102
 
78
103
  ---
79
104
 
80
- ### Command Line Arguments
105
+ **Arguments for `weco run`:**
81
106
 
82
- | Argument | Description | Required |
83
- | :-------------------------- | :----------------------------------------------------------------------------------------------------------------------------------------------------------------------- | :------- |
84
- | `--source` | Path to the source code file that will be optimized (e.g., `optimize.py`). | Yes |
85
- | `--eval-command` | Command to run for evaluating the code in `--source`. This command should print the target `--metric` and its value to the terminal (stdout/stderr). See note below. | Yes |
86
- | `--metric` | The name of the metric you want to optimize (e.g., 'accuracy', 'speedup', 'loss'). This metric name should match what's printed by your `--eval-command`. | Yes |
87
- | `--maximize` | Whether to maximize (`true`) or minimize (`false`) the metric. | Yes |
88
- | `--steps` | Number of optimization steps (LLM iterations) to run. | Yes |
89
- | `--model` | Model identifier for the LLM to use (e.g., `gpt-4o`, `claude-3.7-sonnet`). Recommended models to try include `o4-mini`, and `gemini-2.5-pro-exp-03-25`.| Yes |
90
- | `--additional-instructions` | (Optional) Natural language description of specific instructions OR path to a file containing detailed instructions to guide the LLM. | No |
91
- | `--log-dir` | (Optional) Path to the directory to log intermediate steps and final optimization result. Defaults to `.runs/`. | No |
92
- | `--preserve-source` | (Optional) If set, do not overwrite the original `--source` file. Modifications and the best solution will still be saved in the `--log-dir`. | No |
107
+ | Argument | Description | Required |
108
+ | :-------------------------- | :------------------------------------------------------------------------------------------------------------------------------------------------------------------------ | :------- |
109
+ | `--source` | Path to the source code file that will be optimized (e.g., `optimize.py`). | Yes |
110
+ | `--eval-command` | Command to run for evaluating the code in `--source`. This command should print the target `--metric` and its value to the terminal (stdout/stderr). See note below. | Yes |
111
+ | `--metric` | The name of the metric you want to optimize (e.g., 'accuracy', 'speedup', 'loss'). This metric name should match what's printed by your `--eval-command`. | Yes |
112
+ | `--maximize` | Whether to maximize (`true`) or minimize (`false`) the metric. | Yes |
113
+ | `--steps` | Number of optimization steps (LLM iterations) to run. | Yes |
114
+ | `--model` | Model identifier for the LLM to use (e.g., `gpt-4o`, `claude-3.5-sonnet`). Recommended models to try include `o3-mini`, `claude-3-haiku`, and `gemini-2.5-pro-exp-03-25`. | Yes |
115
+ | `--additional-instructions` | (Optional) Natural language description of specific instructions OR path to a file containing detailed instructions to guide the LLM. | No |
116
+ | `--log-dir` | (Optional) Path to the directory to log intermediate steps and final optimization result. Defaults to `.runs/`. | No |
117
+ | `--preserve-source` | (Optional) If set, do not overwrite the original `--source` file. Modifications and the best solution will still be saved in the `--log-dir`. | No |
118
+
119
+ ---
120
+
121
+ ### `weco logout` Command
122
+
123
+ This command logs you out by removing the locally stored Weco API key.
124
+
125
+ ```bash
126
+ weco logout
127
+ ```
93
128
 
94
129
  ---
95
130
 
@@ -98,6 +133,7 @@ weco --source optimize.py \
98
133
  Weco, powered by the AIDE algorithm, optimizes code iteratively based on your evaluation results. Achieving significant improvements, especially on complex research-level tasks, often requires substantial exploration time.
99
134
 
100
135
  The following plot from the independent [Research Engineering Benchmark (RE-Bench)](https://metr.org/AI_R_D_Evaluation_Report.pdf) report shows the performance of AIDE (the algorithm behind Weco) on challenging ML research engineering tasks over different time budgets.
136
+
101
137
  <p align="center">
102
138
  <img src="https://github.com/user-attachments/assets/ff0e471d-2f50-4e2d-b718-874862f533df" alt="RE-Bench Performance Across Time" width="60%"/>
103
139
  </p>
@@ -124,23 +160,25 @@ Final speedup value = 1.5
124
160
 
125
161
  Weco will parse this output to extract the numerical value (1.5 in this case) associated with the metric name ('speedup').
126
162
 
127
-
128
163
  ## Contributing
129
164
 
130
165
  We welcome contributions! To get started:
131
166
 
132
167
  1. **Fork and Clone the Repository:**
168
+
133
169
  ```bash
134
170
  git clone https://github.com/WecoAI/weco-cli.git
135
171
  cd weco-cli
136
172
  ```
137
173
 
138
174
  2. **Install Development Dependencies:**
175
+
139
176
  ```bash
140
177
  pip install -e ".[dev]"
141
178
  ```
142
179
 
143
180
  3. **Create a Feature Branch:**
181
+
144
182
  ```bash
145
183
  git checkout -b feature/your-feature-name
146
184
  ```
{weco-0.2.12 → weco-0.2.14}/pyproject.toml RENAMED
@@ -10,7 +10,7 @@ authors = [
10
10
  ]
11
11
  description = "Documentation for `weco`, a CLI for using Weco AI's code optimizer."
12
12
  readme = "README.md"
13
- version = "0.2.12"
13
+ version = "0.2.14"
14
14
  license = {text = "MIT"}
15
15
  requires-python = ">=3.8"
16
16
  dependencies = ["requests", "rich"]
weco-0.2.14/weco/__init__.py ADDED
@@ -0,0 +1,14 @@
1
+ import os
2
+
3
+ # DO NOT EDIT
4
+ __pkg_version__ = "0.2.14"
5
+ __api_version__ = "v1"
6
+
7
+ __base_url__ = f"https://api.weco.ai/{__api_version__}"
8
+ # If user specifies a custom base URL, use that instead
9
+ if os.environ.get("WECO_BASE_URL"):
10
+ __base_url__ = os.environ.get("WECO_BASE_URL")
11
+
12
+ __dashboard_url__ = "https://dashboard.weco.ai"
13
+ if os.environ.get("WECO_DASHBOARD_URL"):
14
+ __dashboard_url__ = os.environ.get("WECO_DASHBOARD_URL")
{weco-0.2.12 → weco-0.2.14}/weco/api.py RENAMED
@@ -7,8 +7,7 @@ import sys
7
7
 
8
8
  def handle_api_error(e: requests.exceptions.HTTPError, console: rich.console.Console) -> None:
9
9
  """Extract and display error messages from API responses in a structured format."""
10
- error_message = str(e) # Default message
11
- console.print(f"[bold red]Error:[/] {error_message}")
10
+ console.print(f"[bold red]{e.response.json()['detail']}[/]")
12
11
  sys.exit(1)
13
12
 
14
13
 
@@ -24,13 +23,15 @@ def start_optimization_session(
24
23
  search_policy_config: Dict[str, Any],
25
24
  additional_instructions: str = None,
26
25
  api_keys: Dict[str, Any] = {},
26
+ auth_headers: dict = {}, # Add auth_headers
27
27
  timeout: int = 800,
28
28
  ) -> Dict[str, Any]:
29
29
  """Start the optimization session."""
30
30
  with console.status("[bold green]Starting Optimization..."):
31
31
  try:
32
+ # __base_url__ already contains /v1
32
33
  response = requests.post(
33
- f"{__base_url__}/sessions",
34
+ f"{__base_url__}/sessions", # Path is relative to base_url
34
35
  json={
35
36
  "source_code": source_code,
36
37
  "additional_instructions": additional_instructions,
@@ -43,6 +44,7 @@ def start_optimization_session(
43
44
  },
44
45
  "metadata": {"client_name": "cli", "client_version": __pkg_version__, **api_keys},
45
46
  },
47
+ headers=auth_headers, # Add headers
46
48
  timeout=timeout,
47
49
  )
48
50
  response.raise_for_status()
@@ -57,17 +59,20 @@ def evaluate_feedback_then_suggest_next_solution(
57
59
  execution_output: str,
58
60
  additional_instructions: str = None,
59
61
  api_keys: Dict[str, Any] = {},
62
+ auth_headers: dict = {}, # Add auth_headers
60
63
  timeout: int = 800,
61
64
  ) -> Dict[str, Any]:
62
65
  """Evaluate the feedback and suggest the next solution."""
63
66
  try:
67
+ # __base_url__ already contains /v1
64
68
  response = requests.post(
65
- f"{__base_url__}/sessions/{session_id}/suggest",
69
+ f"{__base_url__}/sessions/{session_id}/suggest", # Path is relative to base_url
66
70
  json={
67
71
  "execution_output": execution_output,
68
72
  "additional_instructions": additional_instructions,
69
73
  "metadata": {**api_keys},
70
74
  },
75
+ headers=auth_headers, # Add headers
71
76
  timeout=timeout,
72
77
  )
73
78
  response.raise_for_status()
@@ -77,12 +82,20 @@ def evaluate_feedback_then_suggest_next_solution(
77
82
 
78
83
 
79
84
  def get_optimization_session_status(
80
- console: rich.console.Console, session_id: str, include_history: bool = False, timeout: int = 800
85
+ console: rich.console.Console,
86
+ session_id: str,
87
+ include_history: bool = False,
88
+ auth_headers: dict = {},
89
+ timeout: int = 800, # Add auth_headers
81
90
  ) -> Dict[str, Any]:
82
91
  """Get the current status of the optimization session."""
83
92
  try:
93
+ # __base_url__ already contains /v1
84
94
  response = requests.get(
85
- f"{__base_url__}/sessions/{session_id}", params={"include_history": include_history}, timeout=timeout
95
+ f"{__base_url__}/sessions/{session_id}", # Path is relative to base_url
96
+ params={"include_history": include_history},
97
+ headers=auth_headers, # Add headers
98
+ timeout=timeout,
86
99
  )
87
100
  response.raise_for_status()
88
101
  return response.json()