Adding scripts and workflow for release notes generation (#81)

.github/actions/ai-release-notes/action.yml

@@ -0,0 +1,86 @@
+name: AI Release Notes
+description: Generate AI release notes using git and openai, outputs 'RELEASE_NOTES' and 'OPENAI_PROMPT'
+
+inputs:
+  OPENAI_API_KEY:
+    required: true
+    type: string
+  GHA_PAT:
+    required: true
+    type: string
+  model_name:
+    required: false
+    type: string
+    default: gpt-4o-mini
+  repo_path:
+    required: false
+    type: string
+  custom_prompt:
+    required: false
+    default: ''
+    type: string
+  git_ref:
+    required: false
+    type: string
+    default: ''
+  head_ref:
+    required: false
+    type: string
+    default: main
+  base_ref:
+    required: false
+    type: string
+    default: main
+
+outputs:
+  RELEASE_NOTES:
+    description: "AI generated release notes"
+    value: ${{ steps.ai_release_notes.outputs.RELEASE_NOTES }}
+  OPENAI_PROMPT:
+    description: "Prompt used to generate release notes"
+    value: ${{ steps.ai_prompt.outputs.OPENAI_PROMPT }}
+
+env:
+  GITHUB_REF: ${{ inputs.git_ref == '' && github.event.pull_request.head.ref || inputs.git_ref }}
+  BASE_REF: ${{ inputs.base_ref == '' && github.base_ref || inputs.base_ref }}
+  HEAD_REF: ${{ inputs.head_ref == '' && github.event.pull_request.head.sha || inputs.head_ref }}
+
+runs:
+  using: "composite"
+  steps:
+    - uses: actions/checkout@v4
+      with:
+        repository: ${{ inputs.repo_path }}
+        token: ${{ inputs.GHA_PAT }}
+        ref: ${{ env.GITHUB_REF }}
+        fetch-depth: 0
+
+    - name: Set Workspace
+      shell: bash
+      run: |
+        pip install tiktoken
+        pip install pytz
+
+    # Github outputs: 'OPENAI_PROMPT'
+    - name: Add Git Info to base prompt
+      id: ai_prompt
+      shell: bash
+      env:
+        BASE_REF: ${{ env.BASE_REF }}
+        HEAD_SHA: ${{ env.HEAD_SHA }}
+        PR_TITLE: ${{ github.event.pull_request.title }}
+        PR_BODY: ${{ github.event.pull_request.body }}
+        MODEL_NAME: ${{ inputs.model_name }}
+        CUSTOM_PROMPT: ${{ inputs.custom_prompt }} # Default: ''
+      run: python .github/scripts/release-notes-prompt.py
+
+    # Github outputs: 'RELEASE_NOTES'
+    - name: Generate AI release notes
+      id: ai_release_notes
+      shell: bash
+      env:
+        OPENAI_API_KEY: ${{ inputs.OPENAI_API_KEY }}
+        CUSTOM_PROMPT: ${{ steps.ai_prompt.outputs.OPENAI_PROMPT }}
+        MODEL_NAME: ${{ inputs.model_name }}
+      run: python .github/scripts/ai-release-notes.py
+

.github/scripts/ai-release-notes.py

@@ -0,0 +1,123 @@
+"""
+AI-powered release notes generator that creates concise and informative release notes from git changes.
+
+This script uses OpenAI's API to analyze git changes (summary, diff, and commit log) and generate
+well-formatted release notes in markdown. It focuses on important changes and their impact,
+particularly highlighting new types and schemas while avoiding repetitive information.
+
+Environment Variables Required:
+    OPENAI_API_KEY: OpenAI API key for authentication
+    CHANGE_SUMMARY: Summary of changes made (optional if CUSTOM_PROMPT provided)
+    CHANGE_DIFF: Git diff of changes (optional if CUSTOM_PROMPT provided)
+    CHANGE_LOG: Git commit log (optional if CUSTOM_PROMPT provided)
+    GITHUB_OUTPUT: Path to GitHub output file
+    CUSTOM_PROMPT: Custom prompt to override default (optional)
+"""
+
+import os
+import requests  # type: ignore
+import json
+import tiktoken # type: ignore
+
+OPENAI_API_KEY = os.environ["OPENAI_API_KEY"]
+CHANGE_SUMMARY = os.environ.get('CHANGE_SUMMARY', '')
+CHANGE_DIFF = os.environ.get('CHANGE_DIFF', '')
+CHANGE_LOG = os.environ.get('CHANGE_LOG', '')
+GITHUB_OUTPUT = os.getenv("GITHUB_OUTPUT")
+OPEN_AI_BASE_URL = "https://api.openai.com/v1"
+OPEN_API_HEADERS = {"Authorization": f"Bearer {OPENAI_API_KEY}", "Content-Type": "application/json"}
+CUSTOM_PROMPT = os.environ.get('CUSTOM_PROMPT', '')
+MODEL_NAME = os.environ.get('MODEL_NAME', 'gpt-3.5-turbo-16k')
+
+def num_tokens_from_string(string: str, model_name: str) -> int:
+    """
+    Calculate the number of tokens in a text string for a specific model.
+    
+    Args:
+        string: The input text to count tokens for
+        model_name: Name of the OpenAI model to use for token counting
+    
+    Returns:
+        int: Number of tokens in the input string
+    """
+    encoding = tiktoken.encoding_for_model(model_name)
+    num_tokens = len(encoding.encode(string))
+    return num_tokens
+
+def truncate_to_token_limit(text, max_tokens, model_name):
+    """
+    Truncate text to fit within a maximum token limit for a specific model.
+    
+    Args:
+        text: The input text to truncate
+        max_tokens: Maximum number of tokens allowed
+        model_name: Name of the OpenAI model to use for tokenization
+    
+    Returns:
+        str: Truncated text that fits within the token limit
+    """
+    encoding = tiktoken.encoding_for_model(model_name)
+    encoded = encoding.encode(text)
+    truncated = encoded[:max_tokens]
+    return encoding.decode(truncated)
+
+def generate_release_notes(model_name):
+    """
+    Generate release notes using OpenAI's API based on git changes.
+    
+    Uses the GPT-3.5-turbo model to analyze change summary, commit log, and code diff
+    to generate concise and informative release notes in markdown format. The notes
+    focus on important changes and their impact, with sections for new types/schemas
+    and other updates.
+    
+    Returns:
+        str: Generated release notes in markdown format
+    
+    Raises:
+        requests.exceptions.RequestException: If the OpenAI API request fails
+    """
+    max_tokens = 14000  # Reserve some tokens for the response
+
+    # Truncate inputs if necessary to fit within token limits
+    change_summary = '' if CUSTOM_PROMPT else truncate_to_token_limit(CHANGE_SUMMARY, 1000, model_name)
+    change_log = '' if CUSTOM_PROMPT else truncate_to_token_limit(CHANGE_LOG, 2000, model_name)
+    change_diff = '' if CUSTOM_PROMPT else truncate_to_token_limit(CHANGE_DIFF, max_tokens - num_tokens_from_string(change_summary, model_name) - num_tokens_from_string(change_log, model_name) - 1000, model_name)
+
+    url = f"{OPEN_AI_BASE_URL}/chat/completions"
+
+    # Construct prompt for OpenAI API
+    openai_prompt = CUSTOM_PROMPT if CUSTOM_PROMPT else f"""Based on the following summary of changes, commit log and code diff, please generate concise and informative release notes:
+    Summary of changes:
+    {change_summary}
+    Commit log:
+    {change_log}
+    Code Diff:
+    {json.dumps(change_diff)}
+    """
+
+    data = {
+        "model": model_name,
+        "messages": [{"role": "user", "content": openai_prompt}],
+        "temperature": 0.7,
+        "max_tokens": 1000,
+    }
+
+    print("----------------------------------------------------------------------------------------------------------")
+    print("POST request to OpenAI")
+    print("----------------------------------------------------------------------------------------------------------")
+    ai_response = requests.post(url, headers=OPEN_API_HEADERS, json=data)
+    print(f"Status Code: {str(ai_response.status_code)}")
+    print(f"Response: {ai_response.text}")
+    ai_response.raise_for_status()
+
+    return ai_response.json()["choices"][0]["message"]["content"]
+
+release_notes = generate_release_notes(MODEL_NAME)
+print("----------------------------------------------------------------------------------------------------------")
+print("OpenAI generated release notes")
+print("----------------------------------------------------------------------------------------------------------")
+print(release_notes)
+
+# Write the release notes to GITHUB_OUTPUT
+with open(GITHUB_OUTPUT, "a") as outputs_file:
+    outputs_file.write(f"RELEASE_NOTES<<EOF\n{release_notes}\nEOF")

.github/scripts/release-notes-prompt.py

@@ -1,23 +1,125 @@
-"""
-This script generates a base prompt for OpenAI to create release notes.
-"""
-
-#!/usr/bin/env python3
-
 import os
+import subprocess
+import json
+import re
+import tiktoken # type: ignore
 from datetime import datetime;
 from pytz import timezone
 
 GITHUB_OUTPUT = os.getenv("GITHUB_OUTPUT")
+BASE_REF = os.getenv("BASE_REF", "main")
+HEAD_SHA = os.environ["HEAD_SHA"]
+PR_TITLE = os.environ["PR_TITLE"]
+PR_BODY = os.environ["PR_BODY"]
+EXISTING_NOTES = os.environ.get("EXISTING_NOTES", "null")
+MODEL_NAME = os.environ.get('MODEL_NAME', 'gpt-3.5-turbo-16k')
+CUSTOM_PROMPT = os.environ.get('CUSTOM_PROMPT', '')
 
+def extract_description_section(pr_body):
+    # Find content between ## Description and the next ## or end of text
+    description_match = re.search(r'## Description\s*\n(.*?)(?=\n##|$)', pr_body, re.DOTALL)
+    if description_match:
+        content = description_match.group(1).strip()
+        # Remove the comment line if it exists
+        comment_pattern = r'\[comment\]:.+?\n'
+        content = re.sub(comment_pattern, '', content)
+        return content.strip()
+    return ""
+
+def extract_ellipsis_important(pr_body):
+    # Find content between <!-- ELLIPSIS_HIDDEN --> and <!-- ELLIPSIS_HIDDEN --> that contains [!IMPORTANT]
+    ellipsis_match = re.search(r'<!--\s*ELLIPSIS_HIDDEN\s*-->(.*?)<!--\s*ELLIPSIS_HIDDEN\s*-->', pr_body, re.DOTALL)
+    if ellipsis_match:
+        content = ellipsis_match.group(1).strip()
+        important_match = re.search(r'\[!IMPORTANT\](.*?)(?=\[!|$)', content, re.DOTALL)
+        if important_match:
+            important_text = important_match.group(1).strip()
+            important_text = re.sub(r'^-+\s*', '', important_text)
+            return important_text.strip()
+    return ""
+
+def extract_coderabbit_summary(pr_body):
+    # Find content between ## Summary by CodeRabbit and the next ## or end of text
+    summary_match = re.search(r'## Summary by CodeRabbit\s*\n(.*?)(?=\n##|$)', pr_body, re.DOTALL)
+    return summary_match.group(1).strip() if summary_match else ""
+
+def num_tokens_from_string(string: str, model_name: str) -> int:
+    """
+    Calculate the number of tokens in a text string for a specific model.
+    
+    Args:
+        string: The input text to count tokens for
+        model_name: Name of the OpenAI model to use for token counting
+    
+    Returns:
+        int: Number of tokens in the input string
+    """
+    encoding = tiktoken.encoding_for_model(model_name)
+    num_tokens = len(encoding.encode(string))
+    return num_tokens
+
+def truncate_to_token_limit(text, max_tokens, model_name):
+    """
+    Truncate text to fit within a maximum token limit for a specific model.
+    
+    Args:
+        text: The input text to truncate
+        max_tokens: Maximum number of tokens allowed
+        model_name: Name of the OpenAI model to use for tokenization
+    
+    Returns:
+        str: Truncated text that fits within the token limit
+    """
+    encoding = tiktoken.encoding_for_model(model_name)
+    encoded = encoding.encode(text)
+    truncated = encoded[:max_tokens]
+    return encoding.decode(truncated)
+
+# Extract sections and combine into PR_OVERVIEW
+description = extract_description_section(PR_BODY)
+important = extract_ellipsis_important(PR_BODY)
+summary = extract_coderabbit_summary(PR_BODY)
+
+PR_OVERVIEW = "\n\n".join(filter(None, [description, important, summary]))
+
+# Get git information
+base_sha = subprocess.getoutput(f"git rev-parse origin/{BASE_REF}") if BASE_REF == 'main' else BASE_REF
+diff_overview = subprocess.getoutput(f"git diff {base_sha}..{HEAD_SHA} --name-status | awk '{{print $2}}' | sort | uniq -c | awk '{{print $2 \": \" $1 \" files changed\"}}'")
+git_log = subprocess.getoutput(f"git log {base_sha}..{HEAD_SHA} --pretty=format:'%h - %s (%an)' --reverse | head -n 50")
+git_diff = subprocess.getoutput(f"git diff {base_sha}..{HEAD_SHA} --minimal --abbrev --ignore-cr-at-eol --ignore-space-at-eol --ignore-space-change --ignore-all-space --ignore-blank-lines --unified=0 --diff-filter=ACDMRT")
+
+max_tokens = 14000  # Reserve some tokens for the response
+changes_summary = truncate_to_token_limit(diff_overview, 1000, MODEL_NAME)
+git_logs = truncate_to_token_limit(git_log, 2000, MODEL_NAME)
+changes_diff = truncate_to_token_limit(git_diff, max_tokens - num_tokens_from_string(changes_summary, MODEL_NAME) - num_tokens_from_string(git_logs, MODEL_NAME) - 1000, MODEL_NAME)
+
+# Get today's existing changelog if any
+existing_changelog = EXISTING_NOTES if EXISTING_NOTES != "null" else None
+existing_changelog_text = f"\nAdditional context:\n{existing_changelog}" if existing_changelog else ""
 TODAY = datetime.now(timezone('US/Eastern')).isoformat(sep=' ', timespec='seconds')
 
-BASE_PROMPT = f"""Based on the following 'PR Information', please generate concise and informative release notes to be read by developers.
+BASE_PROMPT = CUSTOM_PROMPT if CUSTOM_PROMPT else f"""Based on the following 'PR Information', please generate concise and informative release notes to be read by developers.
 Format the release notes with markdown, and always use this structure: a descriptive and very short title (no more than 8 words) with heading level 2, a paragraph with a summary of changes (no header), and if applicable, sections for '🚀 New Features & Improvements', '🐛 Bugs Fixed' and '🔧 Other Updates', with heading level 3, skip respectively the sections if not applicable. 
 Finally include the following markdown comment with the PR merged date: <!-- PR_DATE: {TODAY} -->.
 Avoid being repetitive and focus on the most important changes and their impact, discard any mention of version bumps/updates, changeset files, environment variables or syntax updates.
 PR Information:"""
 
+OPENAI_PROMPT = f"""{BASE_PROMPT}
+Git log summary:
+{changes_summary}
+Commit Messages:
+{git_logs}
+PR Title:
+{PR_TITLE}
+PR Overview:
+{PR_OVERVIEW}{existing_changelog_text}
+Code Diff:
+{json.dumps(changes_diff)}"""
+
+print("OpenAI Prompt")
+print("----------------------------------------------------------------")
+print(OPENAI_PROMPT)
+
 # Write the prompt to GITHUB_OUTPUT
 with open(GITHUB_OUTPUT, "a") as outputs_file:
-    outputs_file.write(f"BASE_PROMPT<<EOF\n{BASE_PROMPT}\nEOF")
+    outputs_file.write(f"OPENAI_PROMPT<<EOF\n{OPENAI_PROMPT}\nEOF")

.github/workflows/changeset-ai-releases.yml

@@ -96,11 +96,6 @@ jobs:
           echo "version=$VERSION"
           echo "prev_version=$PREV_VERSION"
 
-      # Generate base prompt for OpenAI, GITHUB_OUTPUT: 'BASE_PROMPT'
-      - name: Release Notes Prompt
-        id: ai_prompt
-        run: python .github/scripts/release-notes-prompt.py
-
       # Get previous version refs, GITHUB_OUTPUT: 'BASE_REF' and 'HEAD_REF'
       - name: Get Previous Version Refs
         id: version_refs
@@ -109,7 +104,7 @@ jobs:
       # Generate release notes using OpenAI if not already edited, GITHUB_OUTPUT: 'RELEASE_NOTES' and 'OPENAI_PROMPT'
       - name: AI Release Notes
         if: ${{ !contains(github.event.pull_request.labels.*.name, 'openai-edited') }}
-        uses: RooVetGit/Roo-GHA/.github/actions/ai-release-notes@main
+        uses: ./.github/actions/ai-release-notes
         id: ai_release_notes
         with:
           GHA_PAT: ${{ secrets.CROSS_REPO_ACCESS_TOKEN }}
@@ -118,7 +113,6 @@ jobs:
           repo_path: ${{ env.REPO_PATH }}
           base_ref: ${{ steps.version_refs.outputs.base_ref }}
           head_ref: ${{ steps.version_refs.outputs.head_ref }}
-          custom_prompt: ${{ steps.ai_prompt.outputs.BASE_PROMPT }}
 
       # Update CHANGELOG.md with AI-generated notes
       - name: Update Changeset Changelog