feat: update LM Studio integration to remove versioning in URLs and enhance build optimization setup

Author: jruokola

CHANGELOG.md

@@ -60,8 +60,8 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 - **Availability checking** via `/models` endpoint before initialization
 - **Configuration options**:
   - Environment variables: `CODEGRAPH_LMSTUDIO_MODEL`, `CODEGRAPH_LMSTUDIO_URL`, `CODEGRAPH_LMSTUDIO_TIMEOUT`, `CODEGRAPH_MAX_CHUNK_TOKENS`
-  - Config file: `embedding.provider = "lmstudio"`, `embedding.lmstudio_url = "http://localhost:1234/v1"`
-  - Default URL: `http://localhost:1234/v1`
+  - Config file: `embedding.provider = "lmstudio"`, `embedding.lmstudio_url = "http://localhost:1234"`
+  - Default URL: `http://localhost:1234` (the `/v1` path is appended automatically)
 - **Feature flag**: `lmstudio` (requires reqwest)
 - **Performance characteristics**: 50 texts/sec throughput, 500ms typical latency, high memory usage (running full model)
 - **Integration**: Seamless integration with `EmbeddingGenerator` factory, automatic provider selection based on config

Cargo.lock

@@ -29,6 +29,65 @@ authors = ["CodeGraph Team"]
 license = "MIT OR Apache-2.0"
 repository = "https://github.com/codegraph/embedding-system"
 
+# Feature groups for common development scenarios
+[workspace.metadata.cargo-features]
+# Minimal build for MCP server development (fastest build times)
+mcp-minimal = [
+    "ai-enhanced",
+    "embeddings-ollama",
+    "codegraph-ai/anthropic"
+]
+
+# MCP development with LM Studio support
+mcp-lmstudio = [
+    "ai-enhanced",
+    "embeddings-ollama",
+    "embeddings-lmstudio",
+    "codegraph-ai/anthropic",
+    "codegraph-ai/openai-compatible"
+]
+
+# Full local stack (no cloud dependencies)
+local-stack = [
+    "daemon",
+    "ai-enhanced",
+    "embeddings-ollama",
+    "embeddings-lmstudio",
+    "embeddings-local",
+    "codegraph-ai/openai-compatible",
+    "autoagents-experimental"
+]
+
+# Cloud-enabled build
+cloud-enabled = [
+    "daemon",
+    "ai-enhanced",
+    "cloud",
+    "embeddings-jina",
+    "codegraph-ai/anthropic",
+    "codegraph-ai/openai-llm",
+    "server-http",
+    "autoagents-experimental"
+]
+
+# Everything (development/testing)
+full = [
+    "daemon",
+    "ai-enhanced",
+    "cloud",
+    "embeddings",
+    "embeddings-local",
+    "embeddings-openai",
+    "embeddings-ollama",
+    "embeddings-jina",
+    "embeddings-lmstudio",
+    "codegraph-ai/anthropic",
+    "codegraph-ai/openai-llm",
+    "codegraph-ai/openai-compatible",
+    "server-http",
+    "autoagents-experimental"
+]
+
 [workspace.dependencies]
 fastrand = "2.0"
 governor = { version = "0.10.1" }
@@ -223,14 +282,25 @@ inherits = "release"
 debug = true
 
 [profile.dev]
-opt-level = 1
+opt-level = 0        # Faster compilation (0 = no optimization)
 debug = true
 incremental = true
+codegen-units = 256  # Parallel codegen for faster builds
 
 [profile.test]
 opt-level = 1
 debug = true
 
+# Super-fast dev profile for rapid iteration
+[profile.fast-dev]
+inherits = "dev"
+opt-level = 0
+incremental = true
+codegen-units = 256
+debug = 1  # Line numbers only, not full debug info
+lto = false
+panic = "unwind"
+
 # Size-optimized release profile for smallest possible binaries
 [profile.release-size]
 inherits = "release"

Cargo.toml

@@ -36,7 +36,7 @@ export CODEGRAPH_EMBEDDING_DIMENSION=1024               # 384, 768, 1024, 1536,
 ```bash
 export CODEGRAPH_EMBEDDING_PROVIDER=lmstudio
 export CODEGRAPH_LMSTUDIO_MODEL=jina-embeddings-v3      # or jina-embeddings-v4, qwen3-embedding-0.6b, nomic-embed-text-v1.5, etc.
-export CODEGRAPH_LMSTUDIO_URL=http://localhost:1234/v1  # Default LM Studio endpoint
+export CODEGRAPH_LMSTUDIO_URL=http://localhost:1234     # Default LM Studio endpoint (the /v1 path is appended automatically)
 export CODEGRAPH_EMBEDDING_DIMENSION=1024               # Auto-detected for 20+ models, or set manually
 ```
 
@@ -310,7 +310,7 @@ Create `~/.codegraph/config.toml`:
 [embedding]
 provider = "lmstudio"
 model = "jinaai/jina-embeddings-v4"
-lmstudio_url = "http://localhost:1234/v1"
+lmstudio_url = "http://localhost:1234"
 dimension = 2048
 
 [llm]
@@ -324,7 +324,7 @@ lmstudio_url = "http://localhost:1234"
 ```bash
 export CODEGRAPH_EMBEDDING_PROVIDER=lmstudio
 export CODEGRAPH_LMSTUDIO_MODEL=jinaai/jina-embeddings-v4
-export CODEGRAPH_LMSTUDIO_URL=http://localhost:1234/v1
+export CODEGRAPH_LMSTUDIO_URL=http://localhost:1234
 export CODEGRAPH_EMBEDDING_DIMENSION=2048
 ```
 

README.md

@@ -8,31 +8,16 @@ description = "AI-powered caching system for CodeGraph vector operations"
 
 [dependencies]
 codegraph-core = { workspace = true }
-ndarray = { workspace = true }
 serde = { workspace = true }
 serde_json = { workspace = true }
-bincode = { workspace = true }
 tracing = { workspace = true }
 async-trait = { workspace = true }
 tokio = { workspace = true }
 parking_lot = { workspace = true }
-backtrace = { workspace = true }
 dashmap = { workspace = true }
-rayon = { workspace = true }
 crossbeam-channel = { workspace = true }
-chrono = { workspace = true }
-futures = { workspace = true }
 sha2 = { workspace = true }
-sysinfo = { workspace = true }
-num_cpus = { workspace = true }
-once_cell = { workspace = true }
-axum = { workspace = true, features = ["ws"] }
-tower-http = { workspace = true, features = ["cors", "fs"] }
-http = { workspace = true }
-tower = { workspace = true }
-tokio-tungstenite = { workspace = true }
 fastrand = { workspace = true }
-futures-util = "0.3"
 
 [dev-dependencies]
 tokio-test = { workspace = true }

crates/codegraph-cache/Cargo.toml

@@ -357,7 +357,7 @@ fn default_embedding_provider() -> String {
     "auto".to_string()
 }
 fn default_lmstudio_url() -> String {
-    "http://localhost:1234/v1".to_string()
+    "http://localhost:1234".to_string()
 }
 fn default_ollama_url() -> String {
     "http://localhost:11434".to_string()

crates/codegraph-core/src/config_manager.rs

@@ -30,7 +30,6 @@ sha2 = { workspace = true }
 toml = { workspace = true }
 rmcp = { workspace = true }
 schemars = { workspace = true }
-once_cell = { workspace = true }
 lru = { workspace = true }
 atty = { workspace = true }
 

crates/codegraph-mcp/Cargo.toml

@@ -22,7 +22,6 @@ napi-derive = "2.16"
 tokio = { workspace = true }
 serde_json = { workspace = true }
 uuid = { workspace = true }
-chrono = { workspace = true }
 tracing = { workspace = true }
 
 # Local crates

crates/codegraph-napi/Cargo.toml

@@ -36,7 +36,7 @@ impl Default for LmStudioEmbeddingConfig {
     fn default() -> Self {
         Self {
             model: "jinaai/jina-embeddings-v3".to_string(),
-            api_base: "http://localhost:1234/v1".to_string(),
+            api_base: "http://localhost:1234".to_string(),
             timeout: Duration::from_secs(60),
             batch_size: 32,
             max_retries: 3,
@@ -186,7 +186,7 @@ impl LmStudioEmbeddingProvider {
 
     /// Call LM Studio embeddings endpoint
     async fn call_embed_endpoint(&self, texts: &[String]) -> Result<Vec<Vec<f32>>> {
-        let url = format!("{}/embeddings", self.config.api_base);
+        let url = format!("{}/v1/embeddings", self.config.api_base.trim_end_matches('/'));
 
         let request = EmbeddingRequest {
             input: texts.to_vec(),

crates/codegraph-vector/src/lmstudio_embedding_provider.rs

@@ -15,7 +15,7 @@ impl Default for LmStudioRerankerConfig {
     fn default() -> Self {
         Self {
             base_url: std::env::var("LMSTUDIO_URL")
-                .unwrap_or_else(|_| "http://localhost:1234/v1".to_string()),
+                .unwrap_or_else(|_| "http://localhost:1234".to_string()),
             api_key: std::env::var("LMSTUDIO_API_KEY").unwrap_or_else(|_| "lm-studio".to_string()),
             model: std::env::var("LMSTUDIO_RERANK_MODEL")
                 .unwrap_or_else(|_| "text-embedding-3-small".to_string()),
@@ -89,7 +89,7 @@ impl LmStudioReranker {
             input: inputs.clone(),
         };
 
-        let url = format!("{}/embeddings", self.config.base_url.trim_end_matches('/'));
+        let url = format!("{}/v1/embeddings", self.config.base_url.trim_end_matches('/'));
         let response = self
             .client
             .post(&url)