Skip to content

Commit 1fe7b90

Browse files
H4LFdotDEVH4LFdotDEV
committed
refactor: Remove deprecated LiteLLM references
LiteLLM model router has been deprecated in favor of direct API calls. This commit removes all LiteLLM references including: - Deleted config/litellm.yaml entirely - Removed LITELLM_MASTER_KEY and LITELLM_SALT_KEY generation - Updated docker-compose to not reference LiteLLM - Cleaned up .env.example (also removed deprecated Hound/LanceDB sections) - Updated CI workflow and test scripts - Updated documentation The tiered memory system now uses: - Hot: Redis (session cache) - Warm: Graphiti/Neo4j (knowledge graph) - Cold: SQLite (metadata, full-text search) - Archive: Obsidian Vault (human-readable notes)
1 parent 42a7080 commit 1fe7b90

12 files changed

Lines changed: 17 additions & 382 deletions

File tree

‎.env.example‎

Lines changed: 0 additions & 53 deletions
Original file line numberDiff line numberDiff line change
@@ -106,35 +106,6 @@ EMBEDDING_PROVIDER=ollama
106106
EMBEDDING_DIMENSIONS=768
107107

108108

109-
# ─────────────────────────────────────────────────────────────────────────────
110-
# LITELLM ROUTING CONFIGURATION
111-
# ─────────────────────────────────────────────────────────────────────────────
112-
#
113-
# LiteLLM provides intelligent routing, fallbacks, and cost tracking.
114-
# Strategy options:
115-
# cost_optimized - Prefer cheapest model that can handle the task
116-
# quality_first - Prefer best model, fall back on failure
117-
# balanced - Balance cost and quality
118-
# local_first - Try local models first, escalate to API
119-
#
120-
# ─────────────────────────────────────────────────────────────────────────────
121-
122-
LITELLM_MASTER_KEY=sk-litellm-...
123-
LITELLM_PORT=4000
124-
LITELLM_LOG_LEVEL=INFO
125-
126-
# Routing strategy
127-
LITELLM_ROUTING_STRATEGY=local_first
128-
129-
# Cost tracking
130-
LITELLM_BUDGET_DAILY=10.00
131-
LITELLM_BUDGET_MONTHLY=200.00
132-
133-
# Token thresholds for model escalation
134-
ESCALATE_AT_TOKENS=4000
135-
ESCALATE_AT_COMPLEXITY=0.7
136-
137-
138109
# ─────────────────────────────────────────────────────────────────────────────
139110
# REDIS (Hot Tier)
140111
# ─────────────────────────────────────────────────────────────────────────────
@@ -180,27 +151,6 @@ LIVEGREP_BUILD_PARALLELISM=4
180151
LIVEGREP_INDEX_ON_SESSION_END=true
181152

182153

183-
# ─────────────────────────────────────────────────────────────────────────────
184-
# HOUND (Project Search)
185-
# ─────────────────────────────────────────────────────────────────────────────
186-
187-
HOUND_BIND_ADDRESS=127.0.0.1:6080
188-
HOUND_POLL_INTERVAL=30000
189-
HOUND_MAX_CONCURRENT_INDEXERS=2
190-
191-
192-
# ─────────────────────────────────────────────────────────────────────────────
193-
# LANCEDB (Semantic Search Vector Store)
194-
# ─────────────────────────────────────────────────────────────────────────────
195-
196-
LANCEDB_PATH=${CLAUDE_CODE_PP_HOME}/memory/semantic.lance
197-
LANCEDB_TABLE_NAME=code_embeddings
198-
199-
# Chunking settings for semantic indexing
200-
CHUNK_SIZE_LINES=150
201-
CHUNK_OVERLAP_LINES=20
202-
203-
204154
# ─────────────────────────────────────────────────────────────────────────────
205155
# OBSIDIAN VAULT (Archive - Human Readable)
206156
# ─────────────────────────────────────────────────────────────────────────────
@@ -263,22 +213,19 @@ DEFAULT_PERMISSION_LEVEL=standard
263213
# FAST_MODEL=llama3.2:3b
264214
# EMBEDDING_MODEL=nomic-embed-text
265215
# EMBEDDING_PROVIDER=ollama
266-
# LITELLM_ROUTING_STRATEGY=local_first
267216
#
268217
# ─── PRESET: Balanced (Local simple, API for complex) ────────────────────────
269218
# EFFICIENCY_PRESET=balanced
270219
# PRIMARY_MODEL=claude-sonnet-4-5-20250929
271220
# FAST_MODEL=llama3.2:3b
272221
# EMBEDDING_MODEL=nomic-embed-text
273-
# LITELLM_ROUTING_STRATEGY=balanced
274222
#
275223
# ─── PRESET: Quality First (Best models, higher cost) ────────────────────────
276224
# EFFICIENCY_PRESET=quality
277225
# PRIMARY_MODEL=claude-opus-4-5-20251101
278226
# FAST_MODEL=claude-haiku-4-5-20251001
279227
# EMBEDDING_MODEL=voyage-code-3
280228
# EMBEDDING_PROVIDER=voyage
281-
# LITELLM_ROUTING_STRATEGY=quality_first
282229
#
283230
# ─────────────────────────────────────────────────────────────────────────────
284231

‎.github/workflows/ci.yml‎

Lines changed: 0 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -178,7 +178,6 @@ jobs:
178178
# Set required environment variables
179179
export NEO4J_PASSWORD=test
180180
export REDIS_PASSWORD=test
181-
export LITELLM_MASTER_KEY=sk-litellm-test
182181
export ANTHROPIC_API_KEY=""
183182
export OPENAI_API_KEY=""
184183

‎CLAUDE.md‎

Lines changed: 2 additions & 25 deletions
Original file line numberDiff line numberDiff line change
@@ -17,8 +17,7 @@ An AI-native development environment that extends Claude Code with persistent me
1717
├─────────────────────────────────────────────────────────────┤
1818
│ Infrastructure │
1919
│ ├── Redis - Hot memory cache (port 6379) │
20-
│ ├── Neo4j - Knowledge graph (port 7687) │
21-
│ └── LiteLLM - Model routing (port 4000) │
20+
│ └── Neo4j - Knowledge graph (port 7474/7687) │
2221
└─────────────────────────────────────────────────────────────┘
2322
```
2423

@@ -162,16 +161,6 @@ The memory system uses automatic tier promotion based on access patterns:
162161

163162
**Automatic Promotion:** Documents accessed 5+ times are promoted to warm tier (Graphiti knowledge graph). Access tracking uses LRU eviction (max 10k entries) with Redis distributed tracking when available.
164163

165-
### LiteLLM Router
166-
167-
Intelligent model routing with cost optimization.
168-
169-
**Configured Models:**
170-
- `claude-opus-4-5-20251101` - Complex reasoning
171-
- `claude-sonnet-4-5-20251101` - Main development
172-
- `gpt-4o` - Fallback/comparison
173-
- `ollama/llama3.2` - Local inference
174-
175164
## Configuration
176165

177166
### Directory Structure
@@ -180,11 +169,9 @@ Intelligent model routing with cost optimization.
180169
~/.claude-code-pp/
181170
├── config/
182171
│ ├── settings.yaml # Main configuration
183-
│ ├── mcp-servers.json # MCP server definitions
184-
│ └── litellm.yaml # Model routing config
172+
│ └── mcp-servers.json # MCP server definitions
185173
├── memory/
186174
│ ├── sqlite/ # Metadata and FTS
187-
│ ├── lancedb/ # Vector embeddings
188175
│ └── vault/ # Obsidian-compatible notes
189176
├── logs/
190177
└── cache/
@@ -242,7 +229,6 @@ docker-compose -f docker/docker-compose.yaml --profile local-llm up -d
242229
|---------|------|---------|
243230
| redis | 6379 | Hot memory cache |
244231
| neo4j | 7474/7687 | Knowledge graph (HTTP/Bolt) |
245-
| litellm | 4000 | Model routing |
246232
| playwright | 9222 | Browser automation (optional) |
247233
| ollama | 11434 | Local LLM (optional) |
248234

@@ -274,15 +260,6 @@ docker-compose -f docker/docker-compose.yaml logs
274260
docker-compose -f docker/docker-compose.yaml restart redis
275261
```
276262

277-
### LiteLLM routing errors
278-
```bash
279-
# Test endpoint
280-
curl http://localhost:4000/health
281-
282-
# Check configured models
283-
curl http://localhost:4000/v1/models
284-
```
285-
286263
## Development
287264

288265
### Building Swift Controller

‎README.md‎

Lines changed: 6 additions & 7 deletions
Original file line numberDiff line numberDiff line change
@@ -10,7 +10,7 @@ Claude Code++ adds enterprise-grade capabilities to Claude Code through MCP (Mod
1010
- **Search MCP** - Multi-layer search (SQLite FTS → livegrep → Graphiti)
1111
- **System Controller** - macOS Accessibility API integration
1212
- **Research Environment** - Voice conversations + webcam whiteboard capture
13-
- **Infrastructure** - Docker-based services for Redis, Neo4j, and model routing
13+
- **Infrastructure** - Docker-based services for Redis and Neo4j
1414

1515
## Architecture
1616

@@ -35,10 +35,10 @@ Claude Code++ adds enterprise-grade capabilities to Claude Code through MCP (Mod
3535
│ └──────────┘ └───────────┘ └───────────┘ └───────────────────┘ │
3636
│ │
3737
│ INFRASTRUCTURE │
38-
│ ┌──────────┐ ┌───────────┐ ┌───────────┐ ┌───────────────────┐ │
39-
│ │ Redis │ │ Neo4j │ │ SQLite │ │ LiteLLM Router │ │
40-
│ │ Cache │ │ Graph DB │ │ Metadata │ │ Model routing │ │
41-
│ └──────────┘ └───────────┘ └───────────┘ └───────────────────┘ │
38+
│ ┌──────────┐ ┌───────────┐ ┌───────────┐ │
39+
│ │ Redis │ │ Neo4j │ │ SQLite │ │
40+
│ │ Cache │ │ Graph DB │ │ Metadata │ │
41+
│ └──────────┘ └───────────┘ └───────────┘ │
4242
│ │
4343
└──────────────────────────────────────────────────────────────────────────────┘
4444
```
@@ -209,8 +209,7 @@ docker-compose -f docker/docker-compose.yaml ps
209209
| Service | Port | Purpose |
210210
|---------|------|---------|
211211
| redis | 6379 | Hot memory cache |
212-
| neo4j | 7687 | Knowledge graph (Graphiti) |
213-
| litellm | 4000 | Model routing |
212+
| neo4j | 7474/7687 | Knowledge graph (Graphiti) |
214213

215214
## Research Environment
216215

‎config/litellm.yaml‎

Lines changed: 0 additions & 159 deletions
This file was deleted.

0 commit comments

Comments
 (0)