[WIP]: Add support for an MCP to analyse and prioritise PRs

.gitignore

@@ -153,3 +153,6 @@ notebooks/all.txt
 notebooks/ci.txt
 
 quickstart_report.json
+
+# MCP server generated dashboards
+mcp_server/dashboards/

mcp_server/HEURISTIC_LLM_INTEGRATION.md

@@ -0,0 +1,374 @@
+# Heuristic + LLM Integration
+
+This readme explains how the PR scoring works
+
+## Overview
+
+The PR review system integrates **heuristic analysis** with **LLM reasoning** to provide smart, context-aware guidance tailored to each PR's complexity and impact. The system includes comprehensive checks for:
+
+- code quality,
+- strategic alignment,
+- team review status,
+- and readiness for merge.
+
+## How It Works
+
+```
+┌─────────────────────────────────────────────────────────────────┐
+│                    PR Analysis Flow                             │
+│                                                                 │
+│  1. GitHub Data           →  PR details, diff, reviews, CI      │
+│                                                                 │
+│  2. Heuristic Analysis    →  Complexity (0-10)                  │
+│                              Impact (0-10)                      │
+│                              Risk Score (0-10)                  │
+│                              Definition of Ready                │
+│                              Internal Review Status             │
+│                              Strategic Goal Alignment           │
+│                                                                 │
+│  3. Pass to LLM           →  LLM sees all heuristic scores      │
+│                              + Original PR data                 |
+│                              + Google Drive context (optional)  │
+│                                                                 │
+│  4. LLM Analysis          →  For SIMPLE PRs:                    │
+│                              - Quick review checklist           │
+│                              - Fast assessment                  │
+│                                                                 |
+│                              For COMPLEX PRs:                   │
+│                              - Detailed review strategy         │
+│                              - Debug checklist                  │
+│                              - What could go wrong?             │
+│                              - Testing recommendations          │
+│                                                                 │
+│  5. Combined Result       →  Heuristic scores + LLM insights    │
+└─────────────────────────────────────────────────────────────────┘
+```
+
+## Heuristic Scoring System
+
+### Priority Score (0-100)
+
+The priority score is calculated from **4 weighted components**:
+
+```
+Priority = Base Score (50%) + Strategic (30%) + Review (15%) + Staleness (5%)
+Maximum: 50 + 30 + 15 + 5 = 100
+```
+
+#### 1. Base Score (0-50 points)
+
+Based on the **Complexity × Impact Matrix**:
+
+| Complexity   | Impact    | Base Score | Category                                             |
+| ------------ | --------- | ---------- | ---------------------------------------------------- |
+| Simple (≤4)  | High (≥7) | **50**     | 🔥 **CRITICAL** - High impact, easy to review        |
+| Simple (≤4)  | Low (\<7) | **40**     | ⚡ **QUICK WIN** - Easy to review and merge          |
+| Complex (>4) | High (≥7) | **35**     | 🎯 **IMPORTANT** - High impact, needs careful review |
+| Complex (>4) | Low (\<7) | **20**     | 📝 **LOW PRIORITY** - Complex with low impact        |
+
+**Rationale:**
+
+- **Simple + High Impact** gets highest priority (50) → Critical fixes that are easy to review
+- **Simple + Low Impact** gets second priority (40) → Quick wins to clear the backlog
+- **Complex + High Impact** gets medium-high (35) → Important but requires time
+- **Complex + Low Impact** gets lowest (20) → Can be deferred
+
+#### 2. Strategic Score (0-30 points)
+
+Aligns PRs with **Q4 strategic goals**:
+
+- **P0 (Must-Have):** +30 points 🔥
+- **P1 (Should-Have):** +20 points 🎯
+- **P2 (Nice-to-Have):** +10 points 📌
+- **Not Aligned:** 0 points
+
+**How it works:**
+
+- PRs must link to a GitHub issue (e.g., `Closes #123`)
+- Issues are associated with strategic goals via `link_issue_to_goal()`
+- The highest priority goal is used for scoring
+
+**Example:**
+
+```python
+# PR links to Issue #456
+# Issue #456 is linked to "Q4-inference-opt" (P0 goal)
+# → PR gets +30 strategic points
+```
+
+#### 3. Review Score (0-15 points)
+
+Tracks internal and external review progress:
+
+**Bonuses:**
+
+- **+15 pts:** Ready for external review (2+ Thunder team approvals) ✅
+- **+10 pts:** Approved and mergeable (alternative path)
+
+**Penalties:**
+
+- **-5 pts:** Changes requested
+- **-10 pts:** Has merge conflicts
+
+**Capped at:** 0-15 points
+
+#### 4. Staleness Score (0-5 points)
+
+Encourages merging old PRs, especially simple ones:
+
+**Simple PRs (complexity ≤ 4):**
+
+- **>90 days old:** +5 pts
+- **>60 days old:** +3 pts
+- **>30 days old:** +2 pts
+
+**Complex PRs (complexity > 4):**
+
+- **>90 days old:** +2 pts
+- **>60 days old:** +1 pt
+- **No activity in 30+ days:** -2 pts (likely stale)
+
+**Capped at:** 0-5 points
+
+### Complexity Score (0-10)
+
+Measures how hard the PR is to review:
+
+**Scoring factors:**
+
+- **File count:**
+  - \>20 files: +3
+  - \>10 files: +2
+  - \>5 files: +1
+- **Lines changed:**
+  - \>1000 lines: +3
+  - \>500 lines: +2
+  - \>100 lines: +1
+- **Simple indicators:** "fix typo", "update doc", "formatting" → -2
+- **Complex indicators:** "refactor", "architecture", "redesign" → +2
+- **Core files:** >3 core/base/engine files → +2
+
+**Categories:**
+
+- **0-3:** 🟢 Simple (formatting, docs, small fixes)
+- **4-6:** 🟡 Moderate (features, refactoring)
+- **7-10:** 🔴 Complex (architecture changes, large refactors)
+
+### Impact Score (0-10)
+
+Measures the PR's importance to the project:
+
+**Starting point:** 5 (medium impact)
+
+**Adjustments:**
+
+- **Security:** High (≥7) → +3, Medium (≥4) → +1
+- **Urgency:** High (≥7) → +2, Medium (≥4) → +1
+- **Breaking changes:** High (≥7) → +2
+- **High-impact labels:** "critical", "blocker", "bug", "security", "performance" → +2
+- **Low-impact labels:** "documentation", "style", "chore" → -2
+- **Approved & mergeable:** +1
+
+**Capped at:** 0-10
+
+### Risk Score (0-10)
+
+Multi-dimensional risk assessment:
+
+#### Breaking Changes Risk (0-10)
+
+- Keywords: "breaking", "deprecat", "remov", "incompatib" → +5
+- API/interface files modified → +2
+- \>20 files changed → +2
+- "breaking" label → +5
+
+#### Security Risk (0-10)
+
+- Keywords: "security", "vulnerability", "auth", "credential" → +7
+- Security-related files → +3
+- "security" label → +8
+
+#### Urgency Risk (0-10)
+
+- Keywords: "block", "critical", "urgent", "hotfix", "bug" → +4
+- \>90 days old → +3
+- Critical labels → +5
+- \>15 comments → +2 (high engagement)
+
+**Overall risk:** Average of the three components
+
+### Definition of Ready
+
+Checks whether a PR meets the **Thunder Team PR Guidelines**:
+
+#### Checklist (5 checks):
+
+1. **✅ Descriptive Title**
+
+   - Length ≥ 20 characters
+   - Starts with capital letter
+   - Example: "Add support for fused Adam optimizer"
+
+1. **✅ Comprehensive Body**
+
+   - Length ≥ 100 characters
+   - Self-contained (no "per the title")
+   - Explains what & why
+
+1. **✅ Linked Issue**
+
+   - Contains "closes #123", "fixes #456", or "resolves #789"
+   - Exception: Small fixes (typos, formatting) can skip this
+
+1. **✅ CI Passing**
+
+   - All CI checks completed successfully
+   - No failing tests
+
+1. **✅ Not Draft**
+
+   - PR is not marked as draft
+
+**Scoring:**
+
+- **Readiness Score:** (checks_passed / 5) × 100
+- **80-100:** Ready with minor issues
+- **60-79:** Needs attention
+- **0-59:** Not ready
+
+### Internal Review Status
+
+Tracks **Thunder team** review progress per PR guidelines:
+
+**Thunder Team Members:**
+
+- crcrpar, kshitij12345, kiya00, riccardofelluga, beverlylytle, mattteochen, shino16
+
+**Review Requirements:**
+
+- **2 team approvals** required before pinging external maintainers
+- **0 change requests** outstanding
+
+**Status Messages:**
+
+- ✅ "Ready - Can ping external maintainers (@lantiga, @t-vi, @KaelanDt)"
+- ⏳ "Needs 1 more team approval (has 1/2)"
+- ⏳ "Needs 2 team approvals (has 0/2)"
+- 🔄 "Has X change request(s) from team - needs resolution"
+
+**Impact on Priority:**
+
+- Ready for external review → **+15 priority points**
+- Changes requested → **-5 priority points**
+
+### Strategic Goal Alignment
+
+Links PRs to **Q4 strategic goals**:
+
+**Goal Priorities:**
+
+- **P0 (Must-Have):** Critical Q4 deliverables
+- **P1 (Should-Have):** Important but not blocking
+- **P2 (Nice-to-Have):** Future improvements
+
+**How it works:**
+
+1. Create strategic goals:
+
+   ```python
+   add_strategic_goal(
+       goal_id="Q4-inference-opt",
+       title="Inference Optimization",
+       priority="P0",
+       description="Optimize inference performance",
+       theme="Performance",
+   )
+   ```
+
+1. Link issues to goals:
+
+   ```python
+   link_issue_to_goal(issue_number=456, goal_id="Q4-inference-opt")
+   ```
+
+1. PRs linking to those issues get alignment score:
+
+   - **P0:** +30 priority points 🔥
+   - **P1:** +20 priority points 🎯
+   - **P2:** +10 priority points 📌
+
+**Benefit:** Ensures Q4 goals are prioritized in PR reviews!
+
+## LLM Integration
+
+### What the LLM Sees
+
+The LLM receives a comprehensive prompt including:
+
+```markdown
+## HEURISTIC ANALYSIS
+
+Our automated system has analyzed this PR:
+
+**Complexity Score:** 8/10 (COMPLEX)
+**Impact Score:** 9/10 (HIGH IMPACT)
+**Priority Score:** 75/100
+
+**Priority Reasoning:**
+🎯 IMPORTANT (Complex + High Impact)
+Complexity 8/10: 25 files changed, 2000 lines changed, refactors core files
+Impact 9/10: high urgency, performance label, high-impact label
+Strategic: 🔥 P0 STRATEGIC GOAL (closes #456) (+30pts)
+Review: ✅ ready for external review (2+ team approvals) (+15pts)
+Staleness: aging 65 days (+2pts)
+Final: 82/100 (Base:35 + Strategic:30 + Review:15 + Staleness:2)
+
+**Review Guidance:** This is a complex change. Pay special attention to
+architecture, testing, and potential side effects.
+
+---
+
+## DEFINITION OF READY
+**Readiness Score:** 80/100
+**Status:** ⚠️ Not ready - 1 check(s) failing
+
+**Failing Checks:**
+- ❌ CI checks failing: test_distributed.yaml
+
+---
+
+## INTERNAL REVIEW STATUS
+✅ Ready - Can ping external maintainers (@lantiga, @t-vi, @KaelanDt)
+- Team Approvals: 2/2
+- Reviewers: kshitij12345, mattteochen
+
+---
+
+## STRATEGIC GOAL ALIGNMENT
+**Aligned:** Yes 🔥
+**Priority:** P0 (Must-Have)
+**Goal:** Q4 Inference Optimization
+**Closes:** #456
+```
+
+### LLM Response Sections
+
+#### For SIMPLE PRs (complexity < 7):
+
+**2 Sections:**
+
+1. **Summary** - What and why
+1. **Risk Assessment** - Breaking changes, security, urgency
+
+#### For COMPLEX PRs (complexity ≥ 7):
+
+**3 Sections:**
+
+1. **Summary** - What and why
+1. **Risk Assessment** - Breaking changes, security, urgency
+1. **Review Checklist & Debugging Guide** ✨
+   - Key areas to review
+   - Potential issues
+   - Testing strategy
+   - Architecture impact
+   - Debug checklist