Merge branch 'feature/model-update/25-09'

Author: beachdweller

.dockerignore

@@ -1,6 +1,8 @@
 *
 !Dockerfile
 !entrypoint.py
-!requirements.txt
-!prompt.py
+!llm_client.py
+!llm_configs.py
 !locale/*.json
+!prompt.py
+!requirements.txt

.github/workflows/build.yml

@@ -1,4 +1,5 @@
-# begin build.yml
+# begin .github/workflows/build.yml
+
 name: build
 
 on:
@@ -25,7 +26,7 @@ jobs:
       uses: astral-sh/setup-uv@v6
       with:
         version: "latest"
-        pyproject-file: "pyproject.toml"
+        version-file: "pyproject.toml"
         python-version: 3.11
         enable-cache: true
 
@@ -40,37 +41,113 @@ jobs:
     - name: Verify that the Docker image for the action builds
       run: docker build . --file Dockerfile
 
-    - name: Integration test
+  integration-tests-action:
+    needs: build
+    runs-on: ubuntu-latest
+    strategy:
+      matrix:
+        include:
+          - model: gemini-2.5-flash
+            secret: GOOGLE_API_KEY
+          - model: grok-code-fast
+            secret: XAI_API_KEY
+          - model: claude-sonnet-4-20250514
+            secret: CLAUDE_API_KEY
+          - model: google/gemma-2-9b-it
+            secret: NVIDIA_NIM_API_KEY
+          - model: sonar
+            secret: PERPLEXITY_API_KEY
+      fail-fast: false
+    steps:
+    - uses: actions/checkout@v5
+
+    - name: Install uv package manager
+      uses: astral-sh/setup-uv@v6
+      with:
+        version: "latest"
+        version-file: "pyproject.toml"
+        python-version: 3.11
+        enable-cache: true
+
+    - run: uv sync
+
+    - name: Integration test ${{ matrix.model }}
       id: integration
       uses: ./
       with:
         report-files: tests/sample_report.json,tests/json_dict_div_zero_try_except.json
         student-files: tests/sample_code.py
-        api-key: ${{ secrets.GOOGLE_API_KEY }}
+        api-key: ${{ secrets[matrix.secret] }}
         readme-path: tests/sample_readme.md
+        model: ${{ matrix.model }}
         explanation-in: Korean
         fail-expected: True
       timeout-minutes: 5
+    - name: Output and Verify
+      run: |
+        echo "${{ matrix.model }} ${{ steps.integration.outputs.feedback }}"
+        echo "feedback_${{ matrix.model }}=${{ steps.integration.outputs.feedback }}" >> $GITHUB_OUTPUT
+    - name: Verify integration test results
+      run: |
+        . .venv/bin/activate
+        uv pip list
+        echo "feedback_${{ matrix.model }}=${{ steps.integration.outputs.feedback }}" >> $GITHUB_OUTPUT
+        python3 -u -m pytest tests/integration.py
 
-    - name: Integration test XAI
-      id: integration-xai
-      uses: ./
+  integration-tests-env:
+    needs: build
+    runs-on: ubuntu-latest
+    strategy:
+      matrix:
+        model: [
+          gemini-2.5-flash,
+          grok-code-fast,
+          claude-sonnet-4-20250514,
+          google/gemma-2-9b-it,
+          sonar
+        ]
+      fail-fast: false
+    steps:
+    - uses: actions/checkout@v5
+
+    - name: Install uv package manager
+      uses: astral-sh/setup-uv@v6
       with:
-        report-files: tests/sample_report.json,tests/json_dict_div_zero_try_except.json
-        student-files: tests/sample_code.py
-        api-key: ${{ secrets.XAI_API_KEY }}
-        readme-path: tests/sample_readme.md
-        model: grok
+        version: "latest"
+        version-file: "pyproject.toml"
+        python-version: 3.11
+        enable-cache: true
+
+    - run: uv sync
+
+    - name: Integration test ${{ matrix.model }}
+      id: integration
+      env:
+        INPUT_REPORT-FILES: tests/sample_report.json,tests/json_dict_div_zero_try_except.json
+        INPUT_STUDENT-FILES: tests/sample_code.py
+        INPUT_README-PATH: tests/sample_readme.md
+        INPUT_CLAUDE_API_KEY: ${{ secrets.CLAUDE_API_KEY }}
+        INPUT_GEMINI-API-KEY: ${{ secrets.GOOGLE_API_KEY }}
+        INPUT_GROK-API-KEY: ${{ secrets.XAI_API_KEY }}
+        INPUT_NVIDIA-API-KEY: ${{ secrets.NVIDIA_NIM_API_KEY }}
+        INPUT_PERPLEXITY-API-KEY: ${{ secrets.PERPLEXITY_API_KEY }}
+        INPUT_MODEL: ${{ matrix.model }}
         explanation-in: Korean
         fail-expected: True
+      run: |
+        . .venv/bin/activate
+        uv pip list
+        python3 entrypoint.py
       timeout-minutes: 5
-
-    - name: Output the outputs of the integration test of the action
+    - name: Output and Verify
       run: |
-        echo "${{ steps.integration.outputs.feedback }}"
-
+        echo "${{ matrix.model }} ${{ steps.integration.outputs.feedback }}"
+        echo "feedback_${{ matrix.model }}=${{ steps.integration.outputs.feedback }}" >> $GITHUB_OUTPUT
     - name: Verify integration test results
       run: |
-        echo "feedback=${{ steps.integration.outputs.feedback }}" >> $GITHUB_OUTPUT
+        . .venv/bin/activate
+        uv pip list
+        echo "feedback_${{ matrix.model }}=${{ steps.integration.outputs.feedback }}" >> $GITHUB_OUTPUT
         python3 -u -m pytest tests/integration.py
-# end build.yml
+
+# end .github/workflows/build.yml

CHANGELOG.md

@@ -16,6 +16,28 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ### Fixed
 
+## [v0.3.12] - 2025-09-07
+
+### Added
+- Support for multiple LLM providers (Claude, Gemini, Grok, NVIDIA NIM, Perplexity) with dedicated configuration classes in `llm_configs.py`.
+- `llm_client.py` for robust API interaction with retry, timeout, and error handling.
+- Comprehensive integration tests in `build.yml` for multiple models (`gemini-2.5-flash`, `grok-code-fast`, `claude-sonnet-4-20250514`, `google/gemma-2-9b-it`, `sonar`) using both action and environment variable inputs.
+- Model-to-provider mapping in `get_model_key_from_env()` to handle precise model IDs.
+
+### Changed
+- Updated `entrypoint.py` to support flexible model selection with fallback to `gemini-2.5-flash`.
+- Modified `.dockerignore` and `Dockerfile` to include `llm_client.py` and `llm_configs.py`.
+- Updated `action.yml` inputs: `report-files`, `student-files`, `readme-path` now required; `model` defaults to `gemini`; added `fail-expected`.
+- Updated `GeminiConfig` default model to `gemini-2.5-flash` from `gemini-2.0-flash`.
+- Updated `GrokConfig` default model to `grok-code-fast` from `grok-2-1212`.
+- Enhanced `tests/test_entrypoint.py` to align with new fallback model (`gemini-2.5-flash`) and added test for `INPUT_API-KEY`.
+- Improved logging checks in `tests/test_llm_client.py` for robustness.
+- Added file markers in `tests/test_integration.py` and `tests/test_prompt.py`.
+
+### Fixed
+- Resolved 404 errors in integration tests for `google/gemma-2-9b-it` and `sonar` by adding model-to-provider mapping in `get_model_key_from_env()`.
+- Fixed test failures in `test_get_model_key_from_env__fallback_gemini` and `test_get_model_key_from_env__no_model_fallback_gemini` by updating expected model to `gemini-2.5-flash`.
+
 ## [v0.3.7] - 2025-08-01
 
 ### Added

Dockerfile

@@ -16,6 +16,8 @@ FROM cicirello/pyaction:4
 COPY entrypoint.py /entrypoint.py
 COPY requirements.txt /requirements.txt
 COPY prompt.py /prompt.py
+COPY llm_client.py /llm_client.py
+COPY llm_configs.py /llm_configs.py
 COPY locale/ /locale/
 
 RUN python3 -m pip install --upgrade pip

action.yml

@@ -1,44 +1,31 @@
 # begin action.yml
-# Make sure to change the action name here (must be unique).
-# Update the rest as needed.
-name: 'gemini-python-tutor'
-description: 'Gemini Python Tutor for Github Classroom Assignments'
-branding: # Look at documentation for branding options/requirements. These are my favorites.
-  icon: 'book-open'
-  color: 'green'
+name: 'Gemini Python Tutor'
+description: 'Action to generate feedback using LLMs'
 inputs:
   report-files:
-    description: 'Comma-separeted list of json report file paths'
+    description: 'Comma-separated list of test report file paths'
+    required: true
+  student-files:
+    description: 'Comma-separated list of student code file paths'
     required: true
-    default: 'report.json'
-    type: list
   api-key:
-    description: 'API token for AI'
+    description: 'API key for the selected LLM'
     required: true
-    type: string
-  model:  # New input for the model
-    description: 'The Gemini model to use (e.g., gemini-2.0-flash)'
-    required: false
-    default: 'gemini-2.0-flash'
-    type: string
-  student-files:
-    description: "Comma-separated list of student's Python file paths or a glob pattern"
+  model:
+    description: 'LLM model to use (e.g., gemini, grok)'
     required: false
-    default: 'exercise.py'
-    type: list
+    default: 'gemini'
   readme-path:
-    description: 'assignment instructions path'
-    required: false
-    default: 'README.md'
-    type: string
+    description: 'Path to the README file with assignment instructions'
+    required: true
   explanation-in:
-    description: 'Generate explanations in the specified language'
+    description: 'Language for explanations (e.g., English, Korean)'
     required: false
     default: 'English'
-    type: string
-outputs:
-  feedback:
-    description: 'Feedback from the tutor'
+  fail-expected:
+    description: 'Whether test failures are expected (true/false)'
+    required: false
+    default: 'false'
 runs:
   using: 'docker'
   image: 'Dockerfile'

entrypoint.py

@@ -8,11 +8,19 @@
 
 from typing import Any, Dict, Tuple
 
+
+sys.path.insert(
+    0,
+    str(pathlib.Path(__file__).parent.resolve())
+)
+
+
 from llm_client import LLMAPIClient
 from llm_configs import ClaudeConfig, GeminiConfig, GrokConfig, NvidiaNIMConfig, PerplexityConfig
 
 import prompt
 
+
 logging.basicConfig(level=logging.INFO)
 
 
@@ -84,46 +92,69 @@ def get_startwith(key:str, dictionary:dict) -> Any:
 def get_model_key_from_env() -> Tuple[str, str]:
     """
     Extracts the LLM model and API key from environment variables with flexible selection.
+    - Uses INPUT_API-KEY if provided, especially with a specified model.
+    - Falls back to model-specific API keys if INPUT_API-KEY is not set.
     - Raises ValueError if no API keys are available.
-    - Uses single available API key if only one is set.
-    - Prefers specified model's API key if available.
-    - Falls back to Gemini if its API key is available.
+    - Uses model-to-provider mapping for precise model IDs.
     """
     api_key_dict = get_api_key_dict_from_env()
     valid_keys_dict = {k: v for k, v in api_key_dict.items() if v and v.strip()}
 
+    model = os.getenv('INPUT_MODEL', '').lower()
+    general_api_key = os.getenv('INPUT_API-KEY', '').strip()
+
+    # Model-to-provider mapping for precise model IDs
+    model_to_provider = {
+        'google/gemma-2-9b-it': 'nvidia_nim',
+        'sonar': 'perplexity',
+        'gemini-2.5-flash': 'gemini',
+        'grok-code-fast': 'grok',
+        'claude-sonnet-4-20250514': 'claude'
+    }
+
+    # Case 1: Use INPUT_API-KEY if provided
+    if general_api_key:
+        selected_model = model or 'gemini-2.5-flash'  # Default to specific Gemini model
+        logging.info(f"Using INPUT_API-KEY for model: {selected_model}")
+        return selected_model, general_api_key
+
+    # Case 2: No INPUT_API-KEY, check model-specific keys
     if not valid_keys_dict:
         raise ValueError(
             "No API keys provided. Set at least one of:\n"
+            "\tINPUT_API-KEY\n"
             "\tINPUT_CLAUDE_API_KEY\n"
             "\tINPUT_GEMINI-API-KEY\n"
             "\tINPUT_GROK-API-KEY\n"
             "\tINPUT_NVIDIA-API-KEY\n"
             "\tINPUT_PERPLEXITY-API-KEY\n"
         )
 
-    model = os.getenv('INPUT_MODEL', '').lower()
-
-    # Case 1: Only one API key is available
+    # Case 3: Only one API key available
     if len(valid_keys_dict) == 1:
         selected_model, api_key = next(iter(valid_keys_dict.items()))
         logging.info(f"Using single available model: {selected_model}")
         return selected_model, api_key.strip()
 
-    # Case 2: Multiple API keys available
+    # Case 4: Use model-to-provider mapping for specified model
+    provider = model_to_provider.get(model, None)
+    if model and provider and provider in valid_keys_dict:
+        logging.info(f"Using mapped model: {model} with provider: {provider}")
+        return model, valid_keys_dict[provider].strip()
+
+    # Case 5: Fallback to provider-based matching
     if model:
-        # Check if specified model's API key is available
         api_key = get_startwith(model, valid_keys_dict)
         if api_key:
-            logging.info(f"Using specified model: {model}")
+            logging.info(f"Using specified model with provider matching: {model}")
             return model, api_key.strip()
 
-    # Case 3: Fallback to Gemini if available
+    # Case 6: Fallback to Gemini if available
     if 'gemini' in valid_keys_dict:
         logging.info("Falling back to Gemini model")
-        return 'gemini', valid_keys_dict['gemini'].strip()
+        return 'gemini-2.5-flash', valid_keys_dict['gemini'].strip()
 
-    # Case 4: Specified model not available, and Gemini not available
+    # Case 7: No matching model or Gemini
     raise ValueError(
         f"No API key provided for specified model '{model}' and Gemini not available. "
         f"Available models: {', '.join(valid_keys_dict.keys())}"
@@ -150,10 +181,15 @@ def get_config_class_dict() -> Dict[str, type]:
     """
     return {
         'claude': ClaudeConfig,
+        'claude-sonnet-4-20250514': ClaudeConfig,  # Add specific model
         'gemini': GeminiConfig,
+        'gemini-2.5-flash': GeminiConfig,
         'grok': GrokConfig,
+        'grok-code-fast': GrokConfig,
         'nvidia_nim': NvidiaNIMConfig,
+        'google/gemma-2-9b-it': NvidiaNIMConfig,  # Add specific model
         'perplexity': PerplexityConfig,
+        'sonar': PerplexityConfig  # Add specific model
     }