ai-consultants 2.10.1

package/scripts/synthesize.sh

@@ -0,0 +1,332 @@
+#!/bin/bash
+# synthesize.sh - Auto-synthesis engine for AI Consultants
+#
+# Analyzes responses from all consultants and generates an automatic synthesis
+# with consensus score, weighted recommendation, and comparison table.
+#
+# Usage: ./synthesize.sh <responses_dir> <output_file>
+#
+# Requires: claude CLI for synthesis
+
+set -euo pipefail
+
+SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+source "$SCRIPT_DIR/lib/common.sh"
+
+# =============================================================================
+# FALLBACK FUNCTION (defined before use)
+# =============================================================================
+
+generate_fallback_synthesis() {
+    local responses_dir="$1"
+
+    # Calculate basic statistics
+    local total_confidence=0
+    local count=0
+    local consultants_json="[]"
+
+    for f in "$responses_dir"/*.json; do
+        if [[ -f "$f" ]]; then
+            local conf=$(jq -r '.confidence.score // 5' "$f" 2>/dev/null)
+            local name=$(jq -r '.consultant // "unknown"' "$f" 2>/dev/null)
+            total_confidence=$((total_confidence + conf))
+            count=$((count + 1))
+            consultants_json=$(echo "$consultants_json" | jq --arg n "$name" '. + [$n]')
+        fi
+    done
+
+    local avg_confidence=5
+    if [[ $count -gt 0 ]]; then
+        avg_confidence=$((total_confidence / count))
+    fi
+
+    jq -n \
+        --arg timestamp "$(date -Iseconds)" \
+        --argjson count "$count" \
+        --argjson avg_conf "$avg_confidence" \
+        --argjson consultants "$consultants_json" \
+        '{
+            synthesis_version: "2.0-fallback",
+            timestamp: $timestamp,
+            consultants_analyzed: $count,
+            consensus: {
+                score: 50,
+                level: "unknown",
+                description: "Automatic synthesis not available - manual analysis required",
+                agreed_points: [],
+                disagreed_points: []
+            },
+            weighted_recommendation: {
+                approach: "manual_review",
+                summary: "Manual review of responses required",
+                detailed: "Automatic synthesis was not possible. Please consult individual responses.",
+                confidence_weighted_score: $avg_conf,
+                supporting_consultants: $consultants,
+                dissenting_consultants: [],
+                incorporated_insights: []
+            },
+            comparison_table: [],
+            risk_assessment: {
+                overall_risk: "unknown",
+                risks: []
+            },
+            action_items: [
+                {
+                    priority: 1,
+                    action: "Manual review of responses",
+                    rationale: "Automatic synthesis not available"
+                }
+            ],
+            follow_up_questions: [],
+            fallback: true
+        }'
+}
+
+# =============================================================================
+# MAIN SCRIPT
+# =============================================================================
+
+# --- Parameters ---
+RESPONSES_DIR="${1:-}"
+OUTPUT_FILE="${2:-/tmp/synthesis.json}"
+ORIGINAL_QUESTION="${3:-}"
+
+# --- Validation ---
+if [[ -z "$RESPONSES_DIR" || ! -d "$RESPONSES_DIR" ]]; then
+    log_error "Usage: $0 <responses_dir> <output_file> [original_question]"
+    exit 1
+fi
+
+# --- Check for responses ---
+RESPONSE_COUNT=$(find "$RESPONSES_DIR" -name "*.json" -type f 2>/dev/null | head -10 | wc -l)
+if [[ "$RESPONSE_COUNT" -eq 0 ]]; then
+    log_error "No JSON responses found in $RESPONSES_DIR"
+    exit 1
+fi
+
+log_info "Starting automatic synthesis..."
+
+# --- Collect all responses ---
+COMBINED_RESPONSES=""
+CONSULTANTS=()
+CONFIDENCE_SCORES=()
+
+# Use process substitution to handle filenames with spaces correctly
+while IFS= read -r response_file; do
+    if [[ -f "$response_file" && -s "$response_file" ]]; then
+        CONSULTANT=$(jq -r '.consultant // "unknown"' "$response_file" 2>/dev/null)
+        CONFIDENCE=$(jq -r '.confidence.score // 5' "$response_file" 2>/dev/null)
+
+        CONSULTANTS+=("$CONSULTANT")
+        CONFIDENCE_SCORES+=("$CONFIDENCE")
+
+        # Token optimization v2.1: Extract only essential fields instead of full JSON
+        SYNTHESIS_EXTRACT_FIELDS="${SYNTHESIS_EXTRACT_FIELDS:-true}"
+        if [[ "$SYNTHESIS_EXTRACT_FIELDS" == "true" ]]; then
+            SUMMARY=$(jq -r '.response.summary // "N/A"' "$response_file" 2>/dev/null)
+            APPROACH=$(jq -r '.response.approach // "N/A"' "$response_file" 2>/dev/null)
+            PROS=$(jq -r '(.response.pros // []) | join("; ")' "$response_file" 2>/dev/null)
+            CONS=$(jq -r '(.response.cons // []) | join("; ")' "$response_file" 2>/dev/null)
+            CONF_REASONING=$(jq -r '.confidence.reasoning // "N/A"' "$response_file" 2>/dev/null)
+
+            COMBINED_RESPONSES+="
+**$CONSULTANT** (conf:$CONFIDENCE/10, approach:$APPROACH)
+Summary: $SUMMARY
++: $PROS | -: $CONS
+Reasoning: $CONF_REASONING
+---
+"
+        else
+            # Legacy mode: include full JSON
+            COMBINED_RESPONSES+="
+### $CONSULTANT (Confidence: $CONFIDENCE/10)
+$(cat "$response_file")
+
+---
+"
+        fi
+    fi
+done < <(find "$RESPONSES_DIR" -name "*.json" -type f 2>/dev/null | head -10)
+
+NUM_CONSULTANTS=${#CONSULTANTS[@]}
+log_info "Found $NUM_CONSULTANTS responses to synthesize"
+
+# --- Get synthesis strategy ---
+SYNTHESIS_STRATEGY="${SYNTHESIS_STRATEGY:-majority}"
+log_info "Using synthesis strategy: $SYNTHESIS_STRATEGY"
+
+# --- Strategy-specific instructions ---
+get_strategy_instructions() {
+    local strategy="$1"
+
+    case "$strategy" in
+        majority)
+            echo "STRATEGY: MAJORITY - Weight equally, most common wins, focus on consensus" ;;
+        risk_averse)
+            echo "STRATEGY: RISK AVERSE - Prioritize safety, weight risk mentions higher, prefer established solutions, highlight risks" ;;
+        security_first)
+            echo "STRATEGY: SECURITY FIRST - Prioritize security, highlight vulnerabilities, reject insecure recommendations" ;;
+        cost_capped)
+            echo "STRATEGY: COST CAPPED - Prefer simpler solutions, minimize complexity, note cost implications" ;;
+        compare_only)
+            echo "STRATEGY: COMPARE ONLY - No recommendation, present objectively, set approach to 'user_decision_required'" ;;
+        *)
+            echo "STRATEGY: DEFAULT (MAJORITY) - Weight equally, most common wins" ;;
+    esac
+}
+
+STRATEGY_INSTRUCTIONS=$(get_strategy_instructions "$SYNTHESIS_STRATEGY")
+
+# --- Build synthesis prompt (token-optimized v2.2) ---
+SYNTHESIS_PROMPT="You are an expert meta-analyst. Synthesize AI consultant responses.
+Role context: Architect=scalability/design, Pragmatist=simplicity, Advocate=risks/edge-cases, Innovator=creativity.
+
+## Synthesis Strategy
+$STRATEGY_INSTRUCTIONS
+"
+
+if [[ -n "$ORIGINAL_QUESTION" ]]; then
+    SYNTHESIS_PROMPT+="
+## Original Question
+$ORIGINAL_QUESTION
+
+"
+fi
+
+SYNTHESIS_PROMPT+="
+## Consultant Responses
+$COMBINED_RESPONSES
+
+## Instructions
+
+Analyze carefully and produce ONLY valid JSON (no text before or after):
+
+{
+  \"synthesis_version\": \"2.2\",
+  \"timestamp\": \"$(date -Iseconds)\",
+  \"strategy\": \"$SYNTHESIS_STRATEGY\",
+  \"consultants_analyzed\": $NUM_CONSULTANTS,
+  \"consensus\": {
+    \"score\": <0-100>,
+    \"level\": \"<high|medium|low|none>\",
+    \"description\": \"<description of consensus level>\",
+    \"agreed_points\": [\"<points where >= 3 agree>\"],
+    \"disagreed_points\": [
+      {
+        \"topic\": \"<topic>\",
+        \"positions\": {
+          \"<consultant_name>\": \"<position for each consultant that responded, or N/A if they did not address this topic>\"
+        }
+      }
+    ]
+  },
+  \"weighted_recommendation\": {
+    \"approach\": \"<recommended approach>\",
+    \"summary\": \"<summary in 2-3 sentences>\",
+    \"detailed\": \"<detailed explanation>\",
+    \"confidence_weighted_score\": <1-10 weighted by confidence>,
+    \"supporting_consultants\": [\"<who supports>\"],
+    \"dissenting_consultants\": [\"<who dissents>\"],
+    \"incorporated_insights\": [\"<insights from each consultant included>\"]
+  },
+  \"comparison_table\": [
+    {
+      \"aspect\": \"Approach\",
+      \"<consultant_name>\": \"<value for each consultant that responded>\"
+    },
+    {
+      \"aspect\": \"Complexity\",
+      \"<consultant_name>\": \"<value for each consultant that responded>\"
+    },
+    {
+      \"aspect\": \"Scalability\",
+      \"<consultant_name>\": \"<value for each consultant that responded>\"
+    },
+    {
+      \"aspect\": \"Risks\",
+      \"<consultant_name>\": \"<value for each consultant that responded>\"
+    }
+  ],
+  \"risk_assessment\": {
+    \"overall_risk\": \"<low|medium|high>\",
+    \"risks\": [
+      {
+        \"description\": \"<risk>\",
+        \"severity\": \"<low|medium|high>\",
+        \"mitigation\": \"<how to mitigate>\",
+        \"identified_by\": [\"<who identified it>\"]
+      }
+    ]
+  },
+  \"action_items\": [
+    {
+      \"priority\": 1,
+      \"action\": \"<what to do>\",
+      \"rationale\": \"<why>\"
+    }
+  ],
+  \"follow_up_questions\": [\"<questions for further clarification>\"],
+  \"debate_evolution\": {
+    \"initial_positions\": {},
+    \"shifts\": [],
+    \"final_stance\": \"<final stance of the panel>\"
+  }
+}
+
+RULES:
+- consensus.score: 100% = all agree, 75-99% = 3+ agree, 50-74% = 2v2, 25-49% = strong disagreement, 0-24% = no convergence
+- confidence_weighted_score = weighted average by confidence of consultants supporting the approach
+- Include ALL consultants that responded in comparison_table (use consultant name as key, \"N/A\" if no data)
+- For disagreed_points, include all consultants that took a position on that topic
+- Respond ONLY with valid JSON, no markdown or additional text
+"
+
+# --- Execute synthesis ---
+TEMP_OUTPUT=$(mktemp)
+
+# Resolve which CLI to use (avoids invoking agent, walks fallback chain)
+SYNTH_CLI=$(resolve_synthesis_cli 2>/dev/null || echo "")
+
+if [[ -n "$SYNTH_CLI" ]]; then
+    log_info "Running synthesis with $SYNTH_CLI..."
+    build_synthesis_args "$SYNTH_CLI"
+    echo "$SYNTHESIS_PROMPT" | "${SYNTHESIS_ARGS[@]}" > "$TEMP_OUTPUT" 2>/dev/null
+    exit_code=$?
+else
+    log_warn "No synthesis CLI available, using local fallback"
+    # Fallback: generate basic synthesis without LLM
+    generate_fallback_synthesis "$RESPONSES_DIR" > "$TEMP_OUTPUT"
+    exit_code=0
+fi
+
+# --- Post-processing ---
+if [[ $exit_code -eq 0 && -f "$TEMP_OUTPUT" && -s "$TEMP_OUTPUT" ]]; then
+    # Extract only JSON from response (remove any text before/after)
+    RAW_OUTPUT=$(cat "$TEMP_OUTPUT")
+
+    # Try to extract JSON
+    if echo "$RAW_OUTPUT" | jq -e '.' > /dev/null 2>&1; then
+        # It's already valid JSON
+        cat "$TEMP_OUTPUT" > "$OUTPUT_FILE"
+    else
+        # Try to extract JSON from text using sed (portable across macOS/Linux)
+        # This extracts from the first { to the last }
+        JSON_EXTRACTED=$(echo "$RAW_OUTPUT" | sed -n '/{/,/}/p' | tr '\n' ' ' || echo "")
+        if [[ -n "$JSON_EXTRACTED" ]] && echo "$JSON_EXTRACTED" | jq -e '.' > /dev/null 2>&1; then
+            echo "$JSON_EXTRACTED" | jq '.' > "$OUTPUT_FILE"
+        else
+            # Fallback: create minimal structure
+            generate_fallback_synthesis "$RESPONSES_DIR" > "$OUTPUT_FILE"
+        fi
+    fi
+
+    log_success "Synthesis completed: $OUTPUT_FILE"
+    rm -f "$TEMP_OUTPUT"
+    cat "$OUTPUT_FILE"
+else
+    log_error "Synthesis failed"
+    generate_fallback_synthesis "$RESPONSES_DIR" > "$OUTPUT_FILE"
+    rm -f "$TEMP_OUTPUT"
+    cat "$OUTPUT_FILE"
+    exit 1
+fi

package/templates/consultation_report.md

@@ -0,0 +1,323 @@
+# AI Consultation Report v2.0
+
+## Metadata
+
+| Field | Value |
+|-------|-------|
+| **Date** | {{TIMESTAMP}} |
+| **Category** | {{CATEGORY}} |
+| **Consensus Score** | {{CONSENSUS_SCORE}}% ({{CONSENSUS_LEVEL}}) |
+| **Active Consultants** | {{ACTIVE_CONSULTANTS}} |
+| **Debate Rounds** | {{DEBATE_ROUNDS}} |
+
+---
+
+## Question
+
+> {{QUERY}}
+
+---
+
+## Automatic Synthesis
+
+{{#IF_SYNTHESIS}}
+
+### Weighted Recommendation
+
+**Recommended Approach:** {{RECOMMENDED_APPROACH}}
+
+**Confidence-Weighted Score:** {{WEIGHTED_SCORE}}/10
+
+**Summary:** {{SYNTHESIS_SUMMARY}}
+
+### Supporting Consultants
+{{SUPPORTING_CONSULTANTS}}
+
+### Dissenting Consultants
+{{DISSENTING_CONSULTANTS}}
+
+### Points of Agreement
+{{#EACH AGREED_POINTS}}
+- {{.}}
+{{/EACH}}
+
+### Points of Disagreement
+{{#EACH DISAGREED_POINTS}}
+- **{{TOPIC}}**: {{POSITIONS}}
+{{/EACH}}
+
+{{/IF_SYNTHESIS}}
+
+{{#IF_NO_SYNTHESIS}}
+*Automatic synthesis not available. See individual responses.*
+{{/IF_NO_SYNTHESIS}}
+
+---
+
+## Consultant Comparison
+
+| Aspect | Gemini | Codex | Mistral | Kilo |
+|--------|--------|-------|---------|------|
+| **Persona** | The Architect | The Pragmatist | The Devil's Advocate | The Innovator |
+| **Approach** | {{GEMINI_APPROACH}} | {{CODEX_APPROACH}} | {{MISTRAL_APPROACH}} | {{KILO_APPROACH}} |
+| **Confidence** | {{GEMINI_CONFIDENCE}}/10 | {{CODEX_CONFIDENCE}}/10 | {{MISTRAL_CONFIDENCE}}/10 | {{KILO_CONFIDENCE}}/10 |
+
+### Consensus Levels
+
+| Level | Score | Description |
+|-------|-------|-------------|
+| **Unanimous** | 100% | All agree |
+| **High** | 75-99% | Strong agreement |
+| **Medium** | 50-74% | Partial agreement |
+| **Low** | 25-49% | Significant disagreement |
+| **None** | 0-24% | No points of convergence |
+
+---
+
+## Risk Assessment
+
+{{#IF_RISKS}}
+
+**Overall Risk:** {{OVERALL_RISK}}
+
+| Risk | Severity | Mitigation | Identified by |
+|------|----------|------------|---------------|
+{{#EACH RISKS}}
+| {{DESCRIPTION}} | {{SEVERITY}} | {{MITIGATION}} | {{IDENTIFIED_BY}} |
+{{/EACH}}
+
+{{/IF_RISKS}}
+
+---
+
+## Action Items
+
+{{#EACH ACTION_ITEMS}}
+{{PRIORITY}}. **{{ACTION}}** - {{RATIONALE}}
+{{/EACH}}
+
+---
+
+## Individual Responses
+
+### Gemini - The Architect
+
+**Confidence:** {{GEMINI_CONFIDENCE}}/10
+
+**Approach:** {{GEMINI_APPROACH}}
+
+**Summary:**
+> {{GEMINI_SUMMARY}}
+
+{{#IF GEMINI_CODE}}
+**Code:**
+```{{GEMINI_LANGUAGE}}
+{{GEMINI_CODE}}
+```
+{{/IF}}
+
+**Pros:**
+{{#EACH GEMINI_PROS}}
+- {{.}}
+{{/EACH}}
+
+**Cons:**
+{{#EACH GEMINI_CONS}}
+- {{.}}
+{{/EACH}}
+
+<details>
+<summary>Full Response</summary>
+
+{{GEMINI_DETAILED}}
+
+</details>
+
+---
+
+### Codex - The Pragmatist
+
+**Confidence:** {{CODEX_CONFIDENCE}}/10
+
+**Approach:** {{CODEX_APPROACH}}
+
+**Summary:**
+> {{CODEX_SUMMARY}}
+
+{{#IF CODEX_CODE}}
+**Code:**
+```{{CODEX_LANGUAGE}}
+{{CODEX_CODE}}
+```
+{{/IF}}
+
+**Pros:**
+{{#EACH CODEX_PROS}}
+- {{.}}
+{{/EACH}}
+
+**Cons:**
+{{#EACH CODEX_CONS}}
+- {{.}}
+{{/EACH}}
+
+<details>
+<summary>Full Response</summary>
+
+{{CODEX_DETAILED}}
+
+</details>
+
+---
+
+### Mistral - The Devil's Advocate
+
+**Confidence:** {{MISTRAL_CONFIDENCE}}/10
+
+**Approach:** {{MISTRAL_APPROACH}}
+
+**Summary:**
+> {{MISTRAL_SUMMARY}}
+
+{{#IF MISTRAL_CODE}}
+**Code:**
+```{{MISTRAL_LANGUAGE}}
+{{MISTRAL_CODE}}
+```
+{{/IF}}
+
+**Pros:**
+{{#EACH MISTRAL_PROS}}
+- {{.}}
+{{/EACH}}
+
+**Cons:**
+{{#EACH MISTRAL_CONS}}
+- {{.}}
+{{/EACH}}
+
+<details>
+<summary>Full Response</summary>
+
+{{MISTRAL_DETAILED}}
+
+</details>
+
+---
+
+### Kilo - The Innovator
+
+**Confidence:** {{KILO_CONFIDENCE}}/10
+
+**Approach:** {{KILO_APPROACH}}
+
+**Summary:**
+> {{KILO_SUMMARY}}
+
+{{#IF KILO_CODE}}
+**Code:**
+```{{KILO_LANGUAGE}}
+{{KILO_CODE}}
+```
+{{/IF}}
+
+**Pros:**
+{{#EACH KILO_PROS}}
+- {{.}}
+{{/EACH}}
+
+**Cons:**
+{{#EACH KILO_CONS}}
+- {{.}}
+{{/EACH}}
+
+<details>
+<summary>Full Response</summary>
+
+{{KILO_DETAILED}}
+
+</details>
+
+---
+
+{{#IF_DEBATE}}
+
+## Multi-Agent Debate
+
+### Round {{DEBATE_ROUND}}
+
+| Consultant | Position Changed | Confidence Delta | Critiques |
+|------------|------------------|------------------|-----------|
+{{#EACH DEBATE_RESULTS}}
+| {{CONSULTANT}} | {{POSITION_CHANGED}} | {{CONFIDENCE_DELTA}} | {{CRITIQUES_COUNT}} |
+{{/EACH}}
+
+**Stability:** {{DEBATE_STABILITY}}
+
+{{/IF_DEBATE}}
+
+---
+
+## Suggested Follow-ups
+
+{{#EACH FOLLOW_UP_QUESTIONS}}
+- {{.}}
+{{/EACH}}
+
+---
+
+## Output Files
+
+| File | Description |
+|------|-------------|
+| `{{OUTPUT_DIR}}/context.md` | Built context |
+| `{{OUTPUT_DIR}}/gemini.json` | Gemini response |
+| `{{OUTPUT_DIR}}/codex.json` | Codex response |
+| `{{OUTPUT_DIR}}/mistral.json` | Mistral response |
+| `{{OUTPUT_DIR}}/kilo.json` | Kilo response |
+| `{{OUTPUT_DIR}}/voting.json` | Voting report |
+| `{{OUTPUT_DIR}}/synthesis.json` | Automatic synthesis |
+
+---
+
+## Raw Responses
+
+<details>
+<summary>Gemini Raw JSON</summary>
+
+```json
+{{GEMINI_RAW}}
+```
+
+</details>
+
+<details>
+<summary>Codex Raw JSON</summary>
+
+```json
+{{CODEX_RAW}}
+```
+
+</details>
+
+<details>
+<summary>Mistral Raw JSON</summary>
+
+```json
+{{MISTRAL_RAW}}
+```
+
+</details>
+
+<details>
+<summary>Kilo Raw JSON</summary>
+
+```json
+{{KILO_RAW}}
+```
+
+</details>
+
+---
+
+*Report generated by AI Consultants v2.0*