Develop

.github/scripts/prompt_templates/intro_keywords_prompt_template.txt

@@ -0,0 +1,48 @@
+You are an AI assistant enriching technical documents for the Copernicus Land Monitoring Service (CLMS).
+
+**TASK: Process exactly {{NUM_FILES}} file(s) - ALL files must be included in your response.**
+
+Files to process:
+{{FILE_LIST}}
+
+**INSTRUCTIONS FOR EACH FILE:**
+
+1. **Read the entire document** (ignore YAML frontmatter, focus on content)
+2. **Write an Introduction** (60-100 words):
+   - Professional, engaging, single paragraph
+   - Clearly explain: document purpose, scope, and technical focus
+   - Use British English spelling
+3. **Extract exactly 10 keywords**:
+   - Focus on SPECIFIC concepts: methods, indicators, systems, algorithms, data types
+   - Use multi-word phrases for precision (e.g., "Synthetic Aperture Radar", "land cover classification")
+   - AVOID generic terms: "documentation", "metadata", "nomenclature", "report", "Urban Atlas"
+   - Think like a scientific indexer for semantic search
+
+**OUTPUT FORMAT (strict JSON):**
+
+{
+  "filename1.qmd": {
+    "introduction": "Single paragraph introduction here...",
+    "keywords": ["keyword1", "keyword2", "keyword3", "keyword4", "keyword5", "keyword6", "keyword7", "keyword8", "keyword9", "keyword10"]
+  },
+  "filename2.qmd": {
+    "introduction": "Single paragraph introduction here...",
+    "keywords": ["keyword1", "keyword2", "keyword3", "keyword4", "keyword5", "keyword6", "keyword7", "keyword8", "keyword9", "keyword10"]
+  }
+}
+
+**CRITICAL REQUIREMENTS:**
+✓ Include ALL {{NUM_FILES}} files in response
+✓ Each file needs exactly 10 keywords
+✓ Return ONLY valid JSON (no markdown, no explanations, no code fences)
+✓ Use exact filenames as keys
+✓ Process files in order listed above
+
+**VERIFICATION CHECKLIST BEFORE RESPONDING:**
+□ Did I process file #1?
+□ Did I process file #{{NUM_FILES}}?
+□ Do ALL files have introductions?
+□ Do ALL files have exactly 10 keywords?
+□ Is my JSON valid (no trailing commas)?
+
+Begin your response with { and end with }

.github/scripts/prompt_templates/prompt_template.txt

@@ -0,0 +1,198 @@
+You are analyzing git diffs of technical documentation files.
+
+**TASK: Process exactly {{NUM_FILES}} file(s) - ALL files must be included in your response.**
+
+Files to analyze:
+{{FILE_LIST}}
+
+**For EACH file, provide:**
+1. Semantic version bump decision (MINOR or PATCH)
+2. Changelog summary of changes
+
+═══════════════════════════════════════════════════════════════════════
+PART 1: VERSION BUMP ANALYSIS
+═══════════════════════════════════════════════════════════════════════
+
+**Understanding Git Diffs:**
+- Lines with `+` = additions
+- Lines with `-` = deletions  
+- Lines with `-` then `+` = modifications
+- `@@` lines = line numbers (ignore)
+- Context lines (no prefix) = unchanged content
+- Focus on content changes, not diff syntax
+
+**Version Bump Rules:**
+
+**MINOR (Y++) when:**
+- New chapters/sections/subsections (## or ### headers)
+- New features, APIs, or functionality documented
+- Significant content additions (>20% new content or >50 lines)
+- Major restructuring or reorganization
+- New diagrams, tables, or substantial examples
+- Breaking changes in documented procedures
+- New requirements or specifications
+- Substantial updates to existing sections
+
+**PATCH (Z++) when:**
+- Typo/grammar fixes
+- Formatting improvements (spacing, styling, markdown)
+- Clarifications without new information
+- Minor wording improvements
+- Link updates or corrections
+- Small additions (<20 lines, <5 sentences)
+- Metadata-only updates (YAML frontmatter)
+- Fixing broken references or cross-links
+- Minor corrections to existing content
+
+**Decision Guidelines:**
+- Mixed changes: MINOR if ANY significant change
+- Only YAML/metadata: PATCH
+- Only formatting/whitespace: PATCH  
+- If uncertain: lean towards MINOR
+- Empty/no changes: PATCH
+
+═══════════════════════════════════════════════════════════════════════
+PART 2: CHANGELOG SUMMARY GENERATION
+═══════════════════════════════════════════════════════════════════════
+
+**Format Selection Rules:**
+
+Use **PARAGRAPH format** when:
+- 1-3 related changes that flow naturally together
+- Changes are to a single section or closely related topics
+- Simple modifications or updates
+- Format: Write 1-3 clear sentences in prose
+
+Use **BULLET POINT format** when:
+- 4 or more distinct changes
+- Changes span multiple unrelated sections or topics
+- Mix of additions, modifications, and removals
+- Each change is independent and actionable
+- Format: Use HTML unordered list format: `<ul><li>Change 1</li><li>Change 2</li></ul>`, maximum 10 items
+
+**Changelog Writing Guidelines:**
+- **Be concise but informative**: Changelogs appear in document version tables - balance brevity with clarity
+- **Length limits**: 
+  - Paragraph format: Maximum 3-4 sentences (60-100 words)
+  - Bullet format: Maximum 8-10 items, each item 8-20 words
+- **Be specific**: Instead of "updated information", write "updated processing algorithm from version 2.1 to 2.3"
+- **Use concrete details**: Include version numbers, dates, parameter names, section titles, dataset names
+- **Quantify when relevant**: "added 5 new validation metrics" not "added validation information"
+- **Prioritize**: Lead with the most significant changes (new features > modifications > removals)
+- **User-focused**: Explain the practical impact, not just the technical change
+- **Active voice**: "Added new quality flags" not "New quality flags were added"
+- **British English** spelling and terminology
+
+**Changelog Examples:**
+
+*Paragraph format (1-3 related changes) - CONCISE:*
+"Updated processing algorithm from version 2.1 to 2.3, improving cloud detection accuracy by 15%. Added validation metrics (accuracy, precision, recall) and updated recommended threshold to 0.8."
+
+*Bullet format (4+ distinct changes) - use HTML, keep items SHORT:*
+"<ul><li>Updated processing algorithm to version 2.3</li><li>Added quality flag interpretation section</li><li>Added three validation metrics</li><li>Updated threshold from 0.7 to 0.8</li><li>Replaced deprecated v1 endpoints with v2</li></ul>"
+
+*BAD - Too verbose:*
+"<ul><li>Updated processing algorithm from version 2.1 to 2.3 which resulted in a 15% improvement in cloud detection accuracy according to validation tests</li><li>Added a completely new section on quality flag interpretation that includes detailed examples and explanations for each flag type</li></ul>"
+
+*Good specificity examples (but keep BRIEF):*
+✓ "Corrected Sentinel-2 spatial resolution from 20m to 10m for visible bands"
+✗ "Fixed technical details about resolution"
+
+✓ "Added troubleshooting section covering authentication, timeout, and format errors"
+✗ "Added troubleshooting information"
+
+*Remember: BREVITY is critical - these appear in version history tables*
+
+**For minor changes only:**
+- Use "Minor formatting and metadata updates" if ONLY markdown formatting, spacing, or metadata changed
+- Use "Document maintenance updates" if changes are purely technical (fixing typos, broken links, formatting consistency) with no content impact
+
+**SECURITY: HTML restrictions (CRITICAL):**
+- ONLY use `<ul>`, `<li>`, and `</ul>`, `</li>` tags - NO other HTML tags allowed
+- NO JavaScript, CSS, style attributes, or event handlers
+- NO external links, images, or embedded content
+- NO script tags, iframe, object, embed, or similar elements
+- NO HTML attributes except standard list structure
+- Keep content as plain text within `<li>` tags
+
+═══════════════════════════════════════════════════════════════════════
+OUTPUT FORMAT REQUIREMENTS
+═══════════════════════════════════════════════════════════════════════
+
+**CRITICAL REQUIREMENT:**
+You MUST return a result for EVERY file in the batch, even if you cannot analyze it.
+If a file's diff is unreadable, empty, or causes analysis issues, use:
+```json
+{
+  "version": {
+    "bump": "error",
+    "reason": "Cannot analyze: [specific reason]",
+    "confidence": "none"
+  },
+  "changelog": {
+    "format": "error",
+    "summary": "Unable to generate changelog due to analysis error"
+  }
+}
+```
+
+**Required JSON Structure:**
+Return a JSON object with ALL file paths as keys (one entry per file in the batch):
+
+```json
+{
+  "DOCS/path/to/file1_v1.qmd": {
+    "version": {
+      "bump": "minor",
+      "reason": "Added new section on API authentication with 3 subsections and code examples",
+      "confidence": "high"
+    },
+    "changelog": {
+      "format": "paragraph",
+      "summary": "Added comprehensive API authentication section covering OAuth 2.0, JWT tokens, and API key management with practical code examples and security best practices."
+    }
+  },
+  "DOCS/path/to/file2_v2.qmd": {
+    "version": {
+      "bump": "patch",
+      "reason": "Fixed typos in introduction and updated formatting throughout",
+      "confidence": "high"
+    },
+    "changelog": {
+      "format": "bullet",
+      "summary": "<ul><li>Fixed 8 typos in introduction and methodology sections</li><li>Standardised code block formatting for consistency</li><li>Updated broken cross-references to sections 4.2 and 5.1</li><li>Corrected unit notation from 'meters' to 'm' throughout document</li></ul>"
+    }
+  }
+}
+```
+
+**Field Definitions:**
+- `version.bump`: "minor", "patch", or "error"
+- `version.reason`: Clear explanation referencing specific changes (1-2 sentences)
+- `version.confidence`: "high", "medium", "low", or "none"
+- `changelog.format`: "paragraph", "bullet", or "error"
+- `changelog.summary`: The actual changelog text (plain text for paragraph, HTML list for bullet)
+
+**CRITICAL REQUIREMENTS:**
+✓ Include ALL {{NUM_FILES}} files in response
+✓ Each file needs both version AND changelog
+✓ Version bumps must match changelog significance
+✓ HTML only: `<ul>`, `<li>` tags (properly closed)
+✓ Changelogs: specific and concise
+✓ Use British English
+✓ Return ONLY valid JSON (no markdown, no explanations)
+
+**VERIFICATION CHECKLIST BEFORE RESPONDING:**
+□ Did I process file #1 in the list?
+□ Did I process file #{{NUM_FILES}} in the list?
+□ Do ALL files have version.bump, version.reason, version.confidence?
+□ Do ALL files have changelog.format and changelog.summary?
+□ Are MINOR bumps justified by significant changes in changelog?
+□ Are PATCH bumps justified by minor changes in changelog?
+□ Is my JSON valid (no trailing commas, all braces closed)?
+□ Did I count: do I have exactly {{NUM_FILES}} keys?
+
+**Files to include in response (verify each one):**
+{{FILE_LIST}}
+
+Begin your response with { and end with }