@prmichaelsen/remember-mcp 2.2.1 → 2.3.0

package/agent/milestones/milestone-10-shared-spaces.md

@@ -0,0 +1,169 @@
+# Milestone 10: Shared Spaces & Confirmation Flow
+
+**Goal**: Implement shared memory spaces with token-based confirmation for publishing
+**Duration**: 2-3 weeks
+**Dependencies**: M1-M4 (Foundation, Memory, Relationships, Preferences)
+**Status**: Not Started
+
+---
+
+## Overview
+
+Implement a secure, user-controlled system for publishing memories to shared spaces (like "The Void"). Uses token-based confirmation to ensure explicit user consent before any sensitive operation. Includes discovery tools for searching shared spaces.
+
+This milestone introduces:
+- **Token-based confirmation pattern** for sensitive operations
+- **Shared space collections** (Memory_Void, Memory_Public, etc.)
+- **Publishing workflow** with explicit user confirmation
+- **Discovery tools** for searching shared spaces
+- **Generic confirmation system** extensible to other actions
+
+---
+
+## Deliverables
+
+### 1. Confirmation Token Service
+- Token generation and validation
+- Firestore storage for pending confirmations
+- Token expiry (5 minutes)
+- Status tracking (pending, confirmed, denied, expired)
+- Cleanup for expired tokens
+
+### 2. Space Memory Architecture
+- Space collection schema (Memory_{space_name})
+- SpaceMemory type definitions
+- Author attribution
+- Discovery metadata
+
+### 3. Publishing Tools (5 tools)
+- `remember_publish` - Generate confirmation token
+- `remember_confirm` - Execute any pending action
+- `remember_deny` - Cancel any pending action
+- `remember_search_space` - Search shared spaces
+- `remember_query_space` - RAG queries on shared spaces
+
+### 4. Testing & Documentation
+- Unit tests for token service
+- Unit tests for all 5 tools
+- End-to-end publishing flow test
+- README updates
+- Firestore TTL configuration guide
+
+---
+
+## Success Criteria
+
+- [ ] Token service creates and validates tokens correctly
+- [ ] Tokens expire after 5 minutes
+- [ ] `remember_publish` generates tokens with stored parameters
+- [ ] `remember_confirm` executes publish action successfully
+- [ ] `remember_deny` cancels pending actions
+- [ ] Published memories appear in shared space collections
+- [ ] `remember_search_space` finds published memories
+- [ ] `remember_query_space` answers questions about shared memories
+- [ ] All 5 tools integrated into server.ts and server-factory.ts
+- [ ] TypeScript compiles without errors
+- [ ] All tests passing
+- [ ] Build successful
+- [ ] Firestore TTL policy configured
+
+---
+
+## Key Files to Create
+
+```
+src/
+├── services/
+│   └── confirmation-token.service.ts    # Token management
+├── types/
+│   └── space-memory.ts                  # SpaceMemory types
+├── weaviate/
+│   └── space-schema.ts                  # Space collection schemas
+└── tools/
+    ├── publish.ts                       # remember_publish
+    ├── confirm.ts                       # remember_confirm
+    ├── deny.ts                          # remember_deny
+    ├── search-space.ts                  # remember_search_space
+    └── query-space.ts                   # remember_query_space
+
+tests/
+└── unit/
+    ├── confirmation-token.service.test.ts
+    ├── publish.test.ts
+    ├── confirm.test.ts
+    ├── deny.test.ts
+    ├── search-space.test.ts
+    └── query-space.test.ts
+```
+
+---
+
+## Implementation Tasks
+
+See individual task documents:
+- Task 34: Create Confirmation Token Service
+- Task 35: Create Space Memory Types and Schema
+- Task 36: Implement remember_publish Tool
+- Task 37: Implement remember_confirm Tool
+- Task 38: Implement remember_deny Tool
+- Task 39: Implement remember_search_space Tool
+- Task 40: Implement remember_query_space Tool
+- Task 41: Configure Firestore TTL Policy
+- Task 42: Create Tests for Shared Spaces
+- Task 43: Update Documentation
+
+---
+
+## Architecture Notes
+
+### Token Flow
+```
+1. Agent calls remember_publish(memory_id, target="the_void")
+2. System validates memory, generates token, stores params
+3. Agent asks user for confirmation
+4. User confirms → Agent calls remember_confirm(token)
+5. System validates token, fetches memory fresh, publishes
+6. Memory copied to Memory_the_void collection
+```
+
+### Space Collections
+- **Personal**: `Memory_User_123` (single user, sanitized user_id)
+- **Shared**: `Memory_the_void`, `Memory_public_space` (multi-user, snake_case space IDs)
+- **Consistent naming**: `Memory_{snake_case_id}` pattern
+
+**Naming Convention**:
+- Space IDs are **snake_case** (lowercase with underscores): `the_void`, `public_space`
+- Collection names use snake_case IDs: `Memory_the_void`, `Memory_public_space`
+- Display names can use any case/spaces: "The Void", "Public Space"
+- Conversion: Display name → lowercase → replace spaces with underscores
+- Example: "The Void" → `the_void` → `Memory_the_void` collection
+
+### Security
+- One-time use tokens (deleted after use)
+- 5-minute expiry
+- User ownership verification
+- Fresh data fetch during confirmation
+
+---
+
+## Testing Strategy
+
+1. **Unit Tests**: Token service, each tool independently
+2. **Integration Tests**: Full publish flow (request → confirm → verify)
+3. **E2E Tests**: Multi-user discovery scenarios
+4. **Security Tests**: Token expiry, replay attacks, ownership
+
+---
+
+## Future Extensions
+
+This pattern can be extended to other confirmable actions:
+- `remember_retract` - Unpublish from shared space
+- `remember_delete_permanent` - Permanent deletion with confirmation
+- `remember_share_with_user` - Share with specific user
+- Any other sensitive operation requiring explicit consent
+
+---
+
+**Next Milestone**: M11 - Ghost Profiles & Pseudonymous Identity
+**Blockers**: None (builds on M1-M4)

package/agent/progress.yaml

@@ -2,11 +2,11 @@
 
 project:
   name: remember-mcp
-  version: 2.0.2
+  version: 2.3.0
   started: 2026-02-11
   status: in_progress
   current_milestone: M5
-  last_updated: 2026-02-15
+  last_updated: 2026-02-16
 
 milestones:
   - id: M1
@@ -123,6 +123,19 @@ milestones:
     progress: 0%
     estimated_weeks: 1
     
+  - id: M10
+    name: Shared Spaces & Confirmation Flow
+    status: not_started
+    progress: 0%
+    estimated_weeks: 2-3
+    tasks_completed: 0
+    tasks_total: 10
+    notes: |
+      Token-based confirmation pattern for secure publishing
+      5 new MCP tools for shared spaces
+      Space collections with snake_case naming
+      Firestore TTL for automatic token cleanup
+    
   - id: M7
     name: Trust & Permissions
     status: not_started
@@ -368,9 +381,9 @@ tasks:
 
 documentation:
   design_documents: 23
-  milestone_documents: 9
+  milestone_documents: 10
   pattern_documents: 5
-  task_documents: 23
+  task_documents: 33
 
 progress:
   planning: 100%
@@ -378,6 +391,79 @@ progress:
   overall: 50%
 
 recent_work:
+  - date: 2026-02-16
+    description: Created Milestone 10 and Tasks for Shared Spaces
+    items:
+      - ✅ Created M10: Shared Spaces & Confirmation Flow milestone
+      - ✅ Created Task 34: Confirmation Token Service
+      - ✅ Created Task 35: Space Memory Types and Schema
+      - ✅ Created Task 36: Implement remember_publish
+      - ✅ Created Task 37: Implement remember_confirm
+      - ✅ Created Task 38: Implement remember_deny
+      - ✅ Created Task 39: Implement remember_search_space
+      - ✅ Created Task 40: Implement remember_query_space
+      - ✅ Created Task 41: Configure Firestore TTL
+      - ✅ Created Task 42: Create Tests
+      - ✅ Created Task 43: Update Documentation
+      - ✅ Updated naming convention: snake_case for space IDs
+      - ✅ "The Void" → `the_void` → `Memory_the_void`
+      - ✅ Token-based confirmation pattern documented
+      - ✅ 5 new tools planned (publish, confirm, deny, search_space, query_space)
+      - 📋 Ready to begin M10 implementation
+      - 📋 10 tasks, estimated 30 hours total
+      
+  - date: 2026-02-16
+    description: ACP Initialization Complete - Firebase SDK Reverted
+    items:
+      - ✅ AGENT.md confirmed up-to-date (v1.0.3)
+      - ⚠️  ACP updates available (v1.0.3 with new templates and scripts)
+      - ✅ All agent documentation reviewed and current
+      - ✅ Project status verified: 4 milestones complete (M1-M4: 100%)
+      - ✅ Version 2.2.1 confirmed (latest release)
+      - ✅ Reverted commit 328abd3 (Firebase Admin SDK migration)
+      - ✅ Restored @prmichaelsen/firebase-admin-sdk-v8@2.2.0
+      - ✅ 53 unit tests passing (1 skipped integration test)
+      - ✅ Test coverage: 29.76% overall
+      - ✅ TypeScript compiles without errors
+      - ✅ Build successful
+      - ✅ All 12 MCP tools implemented and working
+      - ✅ Comprehensive error handling framework in place
+      - ✅ Dual export: standalone server + factory for mcp-auth
+      - ✅ Multi-tenant architecture with per-user isolation
+      - ✅ Vector search with Weaviate + metadata with Firestore
+      - ✅ 45 content types with dynamic descriptions
+      - ✅ N-way relationships with bidirectional tracking
+      - ✅ User preferences system (6 categories)
+      - 📋 Ready to begin M5: Template System
+      - 📋 Project structure follows ACP and bootstrap patterns
+      - 📋 Documentation comprehensive and accurate
+      
+  - date: 2026-02-16
+    description: ACP Initialization - Complete Project Context Review
+    items:
+      - ✅ AGENT.md confirmed up-to-date (v1.0.3)
+      - ✅ All agent documentation reviewed and current
+      - ✅ Project status verified: 4 milestones complete (M1-M4: 100%)
+      - ✅ Version 2.2.1 confirmed (latest release)
+      - ✅ 53 unit tests passing (1 skipped integration test)
+      - ✅ Test coverage: 24.68% overall
+      - ✅ TypeScript compiles without errors
+      - ✅ Build successful
+      - ✅ All 12 MCP tools implemented and working
+      - ✅ Comprehensive error handling framework in place
+      - ✅ Dual export: standalone server + factory for mcp-auth
+      - ✅ Multi-tenant architecture with per-user isolation
+      - ✅ Vector search with Weaviate + metadata with Firestore
+      - ✅ 45 content types with dynamic descriptions
+      - ✅ N-way relationships with bidirectional tracking
+      - ✅ User preferences system (6 categories)
+      - ✅ LLM provider abstraction (Bedrock implemented)
+      - ✅ Background job service for async operations
+      - ✅ Production-ready with Cloud Run deployment
+      - 📋 Ready to begin M5: Template System
+      - 📋 Project structure follows ACP and bootstrap patterns
+      - 📋 Documentation comprehensive and accurate
+      
   - date: 2026-02-15
     description: All Production Error Tasks COMPLETED (Tasks 23-25)
     items:

package/agent/scripts/install.sh

@@ -0,0 +1,118 @@
+#!/bin/bash
+
+# Agent Context Protocol (ACP) Installation Script
+# This script sets up the ACP directory structure and template files in a project
+
+set -e
+
+# Colors for output
+RED='\033[0;31m'
+GREEN='\033[0;32m'
+YELLOW='\033[1;33m'
+BLUE='\033[0;34m'
+NC='\033[0m' # No Color
+
+# Repository details
+REPO_URL="https://github.com/prmichaelsen/agent-context-protocol.git"
+BRANCH="mainline"
+
+echo -e "${GREEN}Agent Context Protocol (ACP) Installer${NC}"
+echo "========================================"
+echo ""
+
+# Use current directory
+TARGET_DIR="$(pwd)"
+
+echo "Installing ACP to: $TARGET_DIR"
+echo ""
+
+# Check if agent directory already exists
+if [ -d "$TARGET_DIR/agent" ]; then
+    echo -e "${YELLOW}Note: agent/ directory already exists${NC}"
+    echo "All ACP files will be updated to latest versions."
+    echo ""
+fi
+
+# Create temporary directory
+TEMP_DIR=$(mktemp -d)
+trap "rm -rf $TEMP_DIR" EXIT
+
+echo "Cloning ACP repository..."
+if ! git clone --depth 1 --branch "$BRANCH" "$REPO_URL" "$TEMP_DIR" &>/dev/null; then
+    echo -e "${RED}Error: Failed to clone repository${NC}"
+    echo "Please check your internet connection and try again."
+    exit 1
+fi
+
+echo -e "${GREEN}✓${NC} Repository cloned"
+echo ""
+
+# Create directory structure
+echo "Creating directory structure..."
+mkdir -p "$TARGET_DIR/agent/design"
+mkdir -p "$TARGET_DIR/agent/milestones"
+mkdir -p "$TARGET_DIR/agent/patterns"
+mkdir -p "$TARGET_DIR/agent/tasks"
+mkdir -p "$TARGET_DIR/agent/commands"
+mkdir -p "$TARGET_DIR/agent/scripts"
+
+# Create .gitkeep files
+touch "$TARGET_DIR/agent/design/.gitkeep"
+touch "$TARGET_DIR/agent/milestones/.gitkeep"
+touch "$TARGET_DIR/agent/patterns/.gitkeep"
+touch "$TARGET_DIR/agent/tasks/.gitkeep"
+
+echo -e "${GREEN}✓${NC} Directory structure created"
+echo ""
+
+# Copy files
+echo "Installing ACP files..."
+
+# Copy template files (only .template.md files from these directories)
+find "$TEMP_DIR/agent/design" -maxdepth 1 -name "*.template.md" -exec cp {} "$TARGET_DIR/agent/design/" \;
+find "$TEMP_DIR/agent/milestones" -maxdepth 1 -name "*.template.md" -exec cp {} "$TARGET_DIR/agent/milestones/" \;
+find "$TEMP_DIR/agent/patterns" -maxdepth 1 -name "*.template.md" -exec cp {} "$TARGET_DIR/agent/patterns/" \;
+find "$TEMP_DIR/agent/tasks" -maxdepth 1 -name "*.template.md" -exec cp {} "$TARGET_DIR/agent/tasks/" \;
+
+# Copy command template
+cp "$TEMP_DIR/agent/commands/command.template.md" "$TARGET_DIR/agent/commands/"
+
+# Copy all command files (flat structure with dot notation)
+# Copies files like acp.init.md, acp.status.md, deploy.production.md, etc.
+if [ -d "$TEMP_DIR/agent/commands" ]; then
+    find "$TEMP_DIR/agent/commands" -maxdepth 1 -name "*.*.md" -exec cp {} "$TARGET_DIR/agent/commands/" \;
+fi
+
+# Copy progress template
+cp "$TEMP_DIR/agent/progress.template.yaml" "$TARGET_DIR/agent/"
+
+# Copy AGENT.md
+cp "$TEMP_DIR/AGENT.md" "$TARGET_DIR/"
+
+# Copy scripts
+cp "$TEMP_DIR/agent/scripts/update.sh" "$TARGET_DIR/agent/scripts/"
+cp "$TEMP_DIR/agent/scripts/check-for-updates.sh" "$TARGET_DIR/agent/scripts/"
+cp "$TEMP_DIR/agent/scripts/uninstall.sh" "$TARGET_DIR/agent/scripts/"
+cp "$TEMP_DIR/agent/scripts/version.sh" "$TARGET_DIR/agent/scripts/"
+cp "$TEMP_DIR/agent/scripts/install.sh" "$TARGET_DIR/agent/scripts/"
+chmod +x "$TARGET_DIR/agent/scripts"/*.sh
+
+echo -e "${GREEN}✓${NC} All files installed"
+echo ""
+echo -e "${GREEN}Installation complete!${NC}"
+echo ""
+echo -e "${GREEN}Next steps:${NC}"
+echo "1. Create your requirements document:"
+echo "   cp agent/design/requirements.template.md agent/design/requirements.md"
+echo ""
+echo "2. Define your first milestone:"
+echo "   cp agent/milestones/milestone-1-{title}.template.md agent/milestones/milestone-1-foundation.md"
+echo ""
+echo "3. Initialize progress tracking:"
+echo "   cp agent/progress.template.yaml agent/progress.yaml"
+echo ""
+echo "4. Read AGENT.md for complete documentation"
+echo ""
+echo -e "${BLUE}For AI agents:${NC}"
+echo "Type 'AGENT.md: Initialize' to get started."
+echo ""

package/agent/scripts/update.sh

@@ -43,22 +43,34 @@ echo ""
 
 echo "Updating ACP files..."
 
-# Update template files
-cp "$TEMP_DIR/agent/design/design.template.md" "agent/design/"
-cp "$TEMP_DIR/agent/design/requirements.template.md" "agent/design/"
-cp "$TEMP_DIR/agent/milestones/milestone-1-{title}.template.md" "agent/milestones/"
-cp "$TEMP_DIR/agent/tasks/task-1-{title}.template.md" "agent/tasks/"
-cp "$TEMP_DIR/agent/patterns/pattern.template.md" "agent/patterns/"
-cp "$TEMP_DIR/agent/patterns/bootstrap.template.md" "agent/patterns/"
+# Update template files (only .template.md files from these directories)
+find "$TEMP_DIR/agent/design" -maxdepth 1 -name "*.template.md" -exec cp {} "agent/design/" \;
+find "$TEMP_DIR/agent/milestones" -maxdepth 1 -name "*.template.md" -exec cp {} "agent/milestones/" \;
+find "$TEMP_DIR/agent/patterns" -maxdepth 1 -name "*.template.md" -exec cp {} "agent/patterns/" \;
+find "$TEMP_DIR/agent/tasks" -maxdepth 1 -name "*.template.md" -exec cp {} "agent/tasks/" \;
+
+# Update command template
+mkdir -p "agent/commands"
+cp "$TEMP_DIR/agent/commands/command.template.md" "agent/commands/"
+
+# Update all command files (flat structure with dot notation)
+# Copies files like acp.init.md, acp.status.md, deploy.production.md, etc.
+if [ -d "$TEMP_DIR/agent/commands" ]; then
+    find "$TEMP_DIR/agent/commands" -maxdepth 1 -name "*.*.md" -exec cp {} "agent/commands/" \;
+fi
+
+# Update progress template
 cp "$TEMP_DIR/agent/progress.template.yaml" "agent/"
 
 # Update AGENT.md
 cp "$TEMP_DIR/AGENT.md" "."
 
 # Update scripts
-cp "$TEMP_DIR/scripts/update.sh" "agent/scripts/"
-cp "$TEMP_DIR/scripts/check-for-updates.sh" "agent/scripts/"
-cp "$TEMP_DIR/scripts/uninstall.sh" "agent/scripts/"
+cp "$TEMP_DIR/agent/scripts/update.sh" "agent/scripts/"
+cp "$TEMP_DIR/agent/scripts/check-for-updates.sh" "agent/scripts/"
+cp "$TEMP_DIR/agent/scripts/uninstall.sh" "agent/scripts/"
+cp "$TEMP_DIR/agent/scripts/version.sh" "agent/scripts/"
+cp "$TEMP_DIR/agent/scripts/install.sh" "agent/scripts/"
 chmod +x agent/scripts/*.sh
 
 echo -e "${GREEN}✓${NC} All files updated"

package/agent/scripts/version.sh

@@ -0,0 +1,35 @@
+#!/bin/bash
+
+# ACP Version Check Script
+# Extracts and displays the current ACP version from AGENT.md
+
+set -e
+
+# Colors for output
+GREEN='\033[0;32m'
+BLUE='\033[0;34m'
+NC='\033[0m' # No Color
+
+# Check if AGENT.md exists
+if [ ! -f "AGENT.md" ]; then
+    echo "Error: AGENT.md not found in current directory"
+    echo "This script should be run from your project root where AGENT.md is located."
+    exit 1
+fi
+
+# Extract version from AGENT.md
+VERSION=$(grep -m 1 "^\*\*Version\*\*:" AGENT.md | sed 's/.*: //')
+CREATED=$(grep -m 1 "^\*\*Created\*\*:" AGENT.md | sed 's/.*: //')
+STATUS=$(grep -m 1 "^\*\*Status\*\*:" AGENT.md | sed 's/.*: //')
+
+# Display version information
+echo -e "${BLUE}📦 ACP Version Information${NC}"
+echo ""
+echo "Version: $VERSION"
+echo "Created: $CREATED"
+echo "Status: $STATUS"
+echo ""
+echo -e "${GREEN}✓${NC} ACP is installed"
+echo ""
+echo "To check for updates: ./agent/scripts/check-for-updates.sh"
+echo "To update ACP: ./agent/scripts/update.sh"

package/agent/tasks/task-27-implement-llm-provider-interface.md

@@ -0,0 +1,51 @@
+# Task 27: Implement LLM Provider Interface
+
+**Milestone**: Phase 0 - LLM Provider Abstraction  
+**Estimated Time**: 2-3 hours  
+**Dependencies**: None  
+**Status**: Not Started
+
+---
+
+## Objective
+
+Create the core LLM provider interface and type definitions that will be used by all LLM provider implementations.
+
+## Steps
+
+1. **Create LLM types file**
+   - Create `src/llm/types.ts`
+   - Define `LLMMessage` interface (role, content)
+   - Define `LLMCompletionOptions` interface (temperature, maxTokens, etc.)
+   - Define `LLMCompletionResult` interface (content, usage, finishReason)
+   - Define `LLMProvider` interface with `complete()` and `validateConfig()` methods
+
+2. **Add type exports**
+   - Export all interfaces from `types.ts`
+   - Ensure TypeScript strict mode compatibility
+   - Add JSDoc comments for documentation
+
+3. **Create index file**
+   - Create `src/llm/index.ts`
+   - Re-export all types for easy importing
+
+## Verification
+
+- [ ] TypeScript compiles without errors
+- [ ] All interfaces are properly exported
+- [ ] JSDoc comments are present
+- [ ] Types follow existing project conventions
+- [ ] Can import types: `import type { LLMProvider } from '../llm/types.js'`
+
+## Files to Create
+
+- `src/llm/types.ts` - Core type definitions
+- `src/llm/index.ts` - Barrel export file
+
+## Reference
+
+See [`agent/design/llm-provider-abstraction.md`](../design/llm-provider-abstraction.md) lines 128-162 for interface definitions.
+
+---
+
+**Next Task**: [task-28-implement-llm-provider-factory.md](task-28-implement-llm-provider-factory.md)

package/agent/tasks/task-28-implement-llm-provider-factory.md

@@ -0,0 +1,64 @@
+# Task 28: Implement LLM Provider Factory
+
+**Milestone**: Phase 0 - LLM Provider Abstraction  
+**Estimated Time**: 2-3 hours  
+**Dependencies**: task-27-implement-llm-provider-interface.md  
+**Status**: Not Started
+
+---
+
+## Objective
+
+Create the LLM provider factory that instantiates the correct provider based on configuration and provides a singleton instance.
+
+## Steps
+
+1. **Create factory file**
+   - Create `src/llm/factory.ts`
+   - Import `LLMProvider` interface from `types.ts`
+   - Import config from `../config.js`
+
+2. **Implement provider singleton**
+   - Create `providerInstance` variable (initially null)
+   - Implement `getLLMProvider()` function
+   - Return cached instance if exists
+   - Otherwise create new instance based on `config.llm.provider`
+
+3. **Add provider switching logic**
+   - Switch on provider type (bedrock, openai, anthropic)
+   - Import provider classes dynamically
+   - Call `validateConfig()` on instantiation
+   - Log which provider is being used
+   - Throw error for unsupported providers
+
+4. **Add convenience function**
+   - Implement `completeLLM()` function
+   - Wrapper around `getLLMProvider().complete()`
+   - Makes it easy to call LLM without getting provider first
+
+5. **Add error handling**
+   - Handle missing provider implementations gracefully
+   - Provide clear error messages
+   - Validate configuration before use
+
+## Verification
+
+- [ ] TypeScript compiles without errors
+- [ ] Factory returns singleton instance
+- [ ] Throws error for unsupported providers
+- [ ] Validates configuration on instantiation
+- [ ] Logs provider selection
+- [ ] `completeLLM()` convenience function works
+- [ ] Can be imported: `import { getLLMProvider, completeLLM } from '../llm/factory.js'`
+
+## Files to Create
+
+- `src/llm/factory.ts` - Provider factory and singleton
+
+## Reference
+
+See [`agent/design/llm-provider-abstraction.md`](../design/llm-provider-abstraction.md) lines 328-376 for factory implementation.
+
+---
+
+**Next Task**: [task-29-update-config-for-llm.md](task-29-update-config-for-llm.md)

package/agent/tasks/task-29-update-config-for-llm.md

@@ -0,0 +1,71 @@
+# Task 29: Update Config for LLM Providers
+
+**Milestone**: Phase 0 - LLM Provider Abstraction  
+**Estimated Time**: 1-2 hours  
+**Dependencies**: None (can be done in parallel with task-27/28)  
+**Status**: Not Started
+
+---
+
+## Objective
+
+Update the configuration system to support multiple LLM providers with provider-specific settings.
+
+## Steps
+
+1. **Update config.ts**
+   - Add `llm` configuration section
+   - Add provider selection (`LLM_PROVIDER` env var)
+   - Add model selection (`LLM_MODEL` env var)
+   - Add provider-specific configs:
+     - `openai`: apiKey, orgId
+     - `anthropic`: apiKey
+     - `bedrock`: region, accessKeyId, secretAccessKey, sessionToken
+     - `cohere`: apiKey
+
+2. **Add type definitions**
+   - Define `LLMProvider` type: `'openai' | 'anthropic' | 'bedrock' | 'cohere' | 'custom'`
+   - Export type for use in other files
+
+3. **Set defaults**
+   - Default provider: 'openai'
+   - Default model: 'gpt-4o-mini'
+   - Fallback to empty strings for missing keys
+
+4. **Update .env.example**
+   - Add LLM provider configuration section
+   - Document all environment variables
+   - Provide examples for each provider
+   - Mark optional vs required variables
+
+5. **Update config validation**
+   - Add LLM config to validation output
+   - Mask sensitive keys in logs
+   - Validate required fields per provider
+
+## Verification
+
+- [ ] TypeScript compiles without errors
+- [ ] Config exports `llm` section
+- [ ] All provider configs are present
+- [ ] `.env.example` is updated with examples
+- [ ] Config validation includes LLM settings
+- [ ] Sensitive keys are masked in logs
+- [ ] Can access config: `import { config } from './config.js'`
+- [ ] `config.llm.provider` returns correct type
+
+## Files to Modify
+
+- `src/config.ts` - Add LLM configuration
+- `.env.example` - Add LLM environment variables
+
+## Reference
+
+See [`agent/design/llm-provider-abstraction.md`](../design/llm-provider-abstraction.md):
+- Lines 33-67 for environment variables
+- Lines 75-123 for config structure
+- Lines 508-568 for .env.example
+
+---
+
+**Next Task**: [task-30-implement-bedrock-provider.md](task-30-implement-bedrock-provider.md)