@chimerai/cli 1.2.7 → 1.2.8

package/dist/commands/create.js

@@ -883,6 +883,17 @@ ANTHROPIC_API_KEY=
 AZURE_OPENAI_API_KEY=
 AZURE_OPENAI_ENDPOINT=
 
+# Ollama (local models — leave empty to disable auto-prefix)
+OLLAMA_BASE_URL=http://localhost:11434
+
+# RAG / Embedding settings
+# Model used for generating embeddings (e.g. text-embedding-ada-002, nomic-embed-text)
+DEFAULT_EMBEDDING_MODEL=text-embedding-ada-002
+# Dimension must match the embedding model (OpenAI ada-002=1536, nomic-embed-text=768)
+EMBEDDING_DIMENSION=1536
+# Default chat model for RAG responses (e.g. gpt-3.5-turbo, llama3.2)
+DEFAULT_CHAT_MODEL=gpt-3.5-turbo
+
 `;
     if (features.includes('billing')) {
         envContent += `# Stripe

package/dist/templates/ai-service.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"ai-service.d.ts","sourceRoot":"","sources":["../../src/templates/ai-service.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;GAYG;AASH,MAAM,WAAW,iBAAiB;IAChC,OAAO,EAAE,MAAM,EAAE,CAAC;IAClB,KAAK,EAAE,MAAM,EAAE,CAAC;IAChB,WAAW,EAAE,MAAM,CAAC;IACpB,UAAU,EAAE,MAAM,CAAC;CACpB;AAED,MAAM,WAAW,QAAQ;IACvB,WAAW,EAAE,MAAM,CAAC;IACpB,WAAW,EAAE,MAAM,CAAC;IACpB,YAAY,EAAE,MAAM,EAAE,CAAC;CACxB;AAMD,eAAO,MAAM,SAAS,EAAE,MAAM,CAAC,MAAM,EAAE,QAAQ,CAmD9C,CAAC;AAMF,wBAAgB,cAAc,CAAC,SAAS,EAAE,MAAM,GAAG,iBAAiB,GAAG,IAAI,CAQ1E;AAED,wBAAgB,eAAe,CAAC,SAAS,EAAE,MAAM,EAAE,QAAQ,EAAE,iBAAiB,GAAG,IAAI,CAIpF;AAMD,wBAAgB,sBAAsB,IAAI,MAAM,CAwM/C;AAMD,wBAAgB,oBAAoB,IAAI,MAAM,CAG7C;AAMD,wBAAgB,mBAAmB,IAAI,MAAM,CAqN5C;AAMD,wBAAgB,oBAAoB,IAAI,MAAM,CAwG7C;AAMD,wBAAgB,yBAAyB,IAAI,MAAM,CAiHlD;AAMD,wBAAgB,wBAAwB,IAAI,MAAM,CAyGjD;AAMD,wBAAgB,qBAAqB,IAAI,MAAM,CA6I9C;AAED,wBAAgB,kBAAkB,IAAI,MAAM,CAkJ3C;AAMD,wBAAgB,mBAAmB,IAAI,MAAM,CAuN5C;AAOD,wBAAgB,yBAAyB,IAAI,MAAM,CAyLlD;AAMD;;GAEG;AACH,wBAAgB,qBAAqB,CAAC,OAAO,EAAE,MAAM,EAAE,EAAE,KAAK,GAAE,MAAM,EAAO,GAAG,MAAM,CAkKrF;AAED;;GAEG;AACH,wBAAgB,uBAAuB,CAAC,OAAO,EAAE,MAAM,EAAE,GAAG,MAAM,CA4CjE;AAED;;GAEG;AACH,wBAAgB,uBAAuB,CAAC,OAAO,EAAE,MAAM,EAAE,GAAG,MAAM,CA6OjE;AAED;;GAEG;AACH,wBAAgB,6BAA6B,CAAC,OAAO,EAAE,MAAM,EAAE,EAAE,KAAK,GAAE,MAAM,EAAO,GAAG,MAAM,CAsD7F;AAED;;GAEG;AACH,wBAAgB,2BAA2B,IAAI,MAAM,CAwBpD;AAED;;GAEG;AACH,wBAAgB,uBAAuB,CAAC,OAAO,EAAE,MAAM,EAAE,EAAE,KAAK,GAAE,MAAM,EAAO,GAAG,MAAM,CAuEvF;AAED;;GAEG;AACH,wBAAgB,8BAA8B,IAAI,MAAM,CAoBvD"}
+{"version":3,"file":"ai-service.d.ts","sourceRoot":"","sources":["../../src/templates/ai-service.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;GAYG;AASH,MAAM,WAAW,iBAAiB;IAChC,OAAO,EAAE,MAAM,EAAE,CAAC;IAClB,KAAK,EAAE,MAAM,EAAE,CAAC;IAChB,WAAW,EAAE,MAAM,CAAC;IACpB,UAAU,EAAE,MAAM,CAAC;CACpB;AAED,MAAM,WAAW,QAAQ;IACvB,WAAW,EAAE,MAAM,CAAC;IACpB,WAAW,EAAE,MAAM,CAAC;IACpB,YAAY,EAAE,MAAM,EAAE,CAAC;CACxB;AAMD,eAAO,MAAM,SAAS,EAAE,MAAM,CAAC,MAAM,EAAE,QAAQ,CAmD9C,CAAC;AAMF,wBAAgB,cAAc,CAAC,SAAS,EAAE,MAAM,GAAG,iBAAiB,GAAG,IAAI,CAQ1E;AAED,wBAAgB,eAAe,CAAC,SAAS,EAAE,MAAM,EAAE,QAAQ,EAAE,iBAAiB,GAAG,IAAI,CAIpF;AAMD,wBAAgB,sBAAsB,IAAI,MAAM,CAwM/C;AAMD,wBAAgB,oBAAoB,IAAI,MAAM,CAG7C;AAMD,wBAAgB,mBAAmB,IAAI,MAAM,CAqN5C;AAMD,wBAAgB,oBAAoB,IAAI,MAAM,CAwG7C;AAMD,wBAAgB,yBAAyB,IAAI,MAAM,CAiHlD;AAMD,wBAAgB,wBAAwB,IAAI,MAAM,CAyGjD;AAMD,wBAAgB,qBAAqB,IAAI,MAAM,CA6I9C;AAED,wBAAgB,kBAAkB,IAAI,MAAM,CAuJ3C;AAMD,wBAAgB,mBAAmB,IAAI,MAAM,CA4Q5C;AAOD,wBAAgB,yBAAyB,IAAI,MAAM,CAyLlD;AAMD;;GAEG;AACH,wBAAgB,qBAAqB,CAAC,OAAO,EAAE,MAAM,EAAE,EAAE,KAAK,GAAE,MAAM,EAAO,GAAG,MAAM,CAkMrF;AAED;;GAEG;AACH,wBAAgB,uBAAuB,CAAC,OAAO,EAAE,MAAM,EAAE,GAAG,MAAM,CAsFjE;AAED;;GAEG;AACH,wBAAgB,uBAAuB,CAAC,OAAO,EAAE,MAAM,EAAE,GAAG,MAAM,CA6OjE;AAED;;GAEG;AACH,wBAAgB,6BAA6B,CAAC,OAAO,EAAE,MAAM,EAAE,EAAE,KAAK,GAAE,MAAM,EAAO,GAAG,MAAM,CAsD7F;AAED;;GAEG;AACH,wBAAgB,2BAA2B,IAAI,MAAM,CAwBpD;AAED;;GAEG;AACH,wBAAgB,uBAAuB,CAAC,OAAO,EAAE,MAAM,EAAE,EAAE,KAAK,GAAE,MAAM,EAAO,GAAG,MAAM,CAuEvF;AAED;;GAEG;AACH,wBAAgB,8BAA8B,IAAI,MAAM,CAoBvD"}

package/dist/templates/ai-service.js

@@ -1026,7 +1026,12 @@ from typing import List, Optional
 import structlog
 
 from config import settings
-from services.vector_store import vector_store, FAISS_AVAILABLE
+from services import vector_store as _vs_module
+from services.vector_store import FAISS_AVAILABLE
+
+
+def _get_vector_store():
+    return _vs_module.vector_store
 from services.chat_service import chat_service
 from models import ChatCompletionRequest, ChatMessage, MessageRole
 
@@ -1037,7 +1042,7 @@ class RAGService:
     """Service for Retrieval Augmented Generation (RAG)."""
 
     def _check_availability(self):
-        if not FAISS_AVAILABLE or vector_store is None:
+        if not FAISS_AVAILABLE or _get_vector_store() is None:
             raise RuntimeError(
                 "FAISS vector store is not available. "
                 "Install faiss-cpu: pip install faiss-cpu numpy"
@@ -1050,13 +1055,13 @@ class RAGService:
     ) -> dict:
         try:
             self._check_availability()
-            ids = await vector_store.add_texts(documents, metadatas)
+            ids = await _get_vector_store().add_texts(documents, metadatas)
             logger.info("documents_added", count=len(ids))
             return {
                 "status": "success",
                 "added": len(ids),
                 "ids": ids,
-                "total_vectors": vector_store.get_stats()["total_vectors"],
+                "total_vectors": _get_vector_store().get_stats()["total_vectors"],
             }
         except Exception as e:
             logger.error("add_documents_failed", error=str(e))
@@ -1065,7 +1070,7 @@ class RAGService:
     async def search_documents(self, query: str, k: int = 4) -> List[dict]:
         try:
             self._check_availability()
-            results = await vector_store.similarity_search(query, k=k)
+            results = await _get_vector_store().similarity_search(query, k=k)
             logger.info("documents_searched", query_length=len(query), results=len(results))
             return results
         except Exception as e:
@@ -1085,7 +1090,7 @@ class RAGService:
         try:
             self._check_availability()
 
-            relevant_docs = await vector_store.similarity_search(query, k=k)
+            relevant_docs = await _get_vector_store().similarity_search(query, k=k)
 
             context_parts = []
             for i, doc in enumerate(relevant_docs, 1):
@@ -1104,7 +1109,7 @@ Context:
                 ChatMessage(role=MessageRole.USER, content=query),
             ]
 
-            model = model or settings.default_chat_model
+            model = model or settings.resolved_chat_model
 
             chat_request = ChatCompletionRequest(
                 model=model,
@@ -1141,22 +1146,22 @@ Context:
 
     def get_stats(self) -> dict:
         self._check_availability()
-        return vector_store.get_stats()
+        return _get_vector_store().get_stats()
 
     def clear_store(self):
         self._check_availability()
-        vector_store.clear()
+        _get_vector_store().clear()
         return {"status": "success", "message": "Vector store cleared"}
 
     async def delete_documents(self, document_ids: List[int]) -> dict:
         try:
             self._check_availability()
-            deleted = await vector_store.delete_by_ids(document_ids)
+            deleted = await _get_vector_store().delete_by_ids(document_ids)
             logger.info("documents_deleted", requested=len(document_ids), deleted=deleted)
             return {
                 "status": "success",
                 "deleted": deleted,
-                "remaining_vectors": vector_store.get_stats()["total_vectors"],
+                "remaining_vectors": _get_vector_store().get_stats()["total_vectors"],
             }
         except Exception as e:
             logger.error("delete_documents_failed", error=str(e))
@@ -1183,6 +1188,7 @@ except Exception as e:
 import pickle
 import os
 from typing import List, Dict, Any, Optional
+import httpx
 import litellm
 from langchain_text_splitters import RecursiveCharacterTextSplitter
 from config import settings
@@ -1345,8 +1351,32 @@ class VectorStore:
 
     async def _generate_embeddings(self, texts: List[str]) -> List[List[float]]:
         try:
+            model = settings.resolved_embedding_model
+            is_ollama = model.startswith("ollama/") or model.startswith("ollama_chat/")
+
+            if is_ollama:
+                # Use Ollama's /api/embed endpoint directly (supports array input)
+                model_name = model.split("/", 1)[1]  # strip "ollama/" prefix
+                base_url = settings.ollama_base_url.rstrip("/")
+                async with httpx.AsyncClient(timeout=60.0) as client:
+                    resp = await client.post(
+                        f"{base_url}/api/embed",
+                        json={"model": model_name, "input": texts},
+                    )
+                    resp.raise_for_status()
+                    data = resp.json()
+                    embeddings = data.get("embeddings") or data.get("embedding")
+                    if embeddings is None:
+                        raise ValueError(f"Unexpected Ollama embed response: {data}")
+                    # /api/embed always returns a list of vectors
+                    if isinstance(embeddings[0], (int, float)):
+                        embeddings = [embeddings]  # single vector → wrap
+                logger.info("embeddings_generated", count=len(embeddings), model=model)
+                return embeddings
+
+            # Non-Ollama: use LiteLLM
             response = await litellm.aembedding(
-                model=settings.default_embedding_model,
+                model=model,
                 input=texts,
             )
             embeddings = [
@@ -1355,7 +1385,7 @@ class VectorStore:
             ]
             logger.info("embeddings_generated",
                         count=len(embeddings),
-                        model=settings.default_embedding_model)
+                        model=settings.resolved_embedding_model)
             return embeddings
         except Exception as e:
             logger.error("embedding_generation_failed", error=str(e))
@@ -1375,14 +1405,42 @@ class VectorStore:
         logger.info("faiss_index_cleared")
 
 
-if FAISS_AVAILABLE:
+# Initialized lazily at startup via init_vector_store()
+vector_store: Optional["VectorStore"] = None
+
+
+def init_vector_store(dimension: int) -> Optional["VectorStore"]:
+    """Initialize (or re-initialize) the global vector store with the given dimension.
+
+    If an existing FAISS index has a different dimension it is automatically
+    cleared and rebuilt — no manual file deletion required.
+    """
+    global vector_store
+
+    if not FAISS_AVAILABLE:
+        logger.warning("faiss_not_available_skipping_vector_store")
+        return None
+
     try:
-        vector_store = VectorStore(dimension=settings.embedding_dimension)
+        instance = VectorStore(dimension=dimension)
+
+        # Dimension mismatch: existing index was built with a different model
+        if instance.index is not None and instance.index.d != dimension:
+            logger.warning(
+                "embedding_dimension_mismatch",
+                index_dimension=instance.index.d,
+                model_dimension=dimension,
+                action="clearing_index",
+            )
+            instance._create_new_index()
+            instance.save()
+
+        vector_store = instance
+        logger.info("vector_store_initialized", dimension=dimension)
+        return vector_store
     except Exception as e:
-        logger.warning(f"Failed to initialize vector store: {e}")
-        vector_store = None
-else:
-    vector_store = None
+        logger.error("vector_store_init_failed", error=str(e))
+        return None
 `;
 }
 // ============================================================================
@@ -1593,6 +1651,7 @@ function generateAiServiceMain(modules, tools = []) {
         routers.push('app.include_router(chat_router)');
     }
     if (hasRag) {
+        imports.push('from services.vector_store import init_vector_store');
         imports.push('from routes.rag_routes import router as rag_router');
         routers.push('app.include_router(rag_router)');
     }
@@ -1657,7 +1716,38 @@ async def lifespan(app: FastAPI):
         except Exception as exc:
             logger.warning("provider_client_error", error=str(exc))
 
-    logger.info("ai_service_started", provider_mode=provider_mode)
+    # --- Auto-detect embedding dimension (RAG) ---
+    detected_dim = settings.embedding_dimension  # fallback from .env / default
+    model = settings.resolved_embedding_model
+    embed_kwargs: dict = {}
+    if model.startswith("ollama/") or model.startswith("ollama_chat/"):
+        embed_kwargs["api_base"] = settings.ollama_base_url
+
+    try:
+        is_ollama = model.startswith("ollama/") or model.startswith("ollama_chat/")
+        if is_ollama:
+            import httpx
+            model_name = model.split("/", 1)[1]
+            base_url = settings.ollama_base_url.rstrip("/")
+            async with httpx.AsyncClient(timeout=30.0) as hx:
+                resp = await hx.post(f"{base_url}/api/embed", json={"model": model_name, "input": ["dimension probe"]})
+                resp.raise_for_status()
+                data = resp.json()
+                vec = (data.get("embeddings") or [data.get("embedding")])[0]
+                detected_dim = len(vec)
+        else:
+            import litellm
+            test_resp = await litellm.aembedding(model=model, input=["dimension probe"], **embed_kwargs)
+            detected_dim = len(test_resp.data[0]["embedding"] if isinstance(test_resp.data[0], dict) else test_resp.data[0].embedding)
+        logger.info("embedding_dimension_detected", model=model, dimension=detected_dim)
+    except Exception as exc:
+        logger.warning("embedding_dimension_detection_failed", model=model, error=str(exc),
+                       fallback_dimension=detected_dim)
+
+    settings.embedding_dimension = detected_dim
+    init_vector_store(detected_dim)
+
+    logger.info("ai_service_started", provider_mode=provider_mode, embedding_model=model, embedding_dimension=detected_dim)
     yield
 
     await provider_client.close()
@@ -1751,13 +1841,36 @@ function generateAiServiceConfig(modules) {
     # RAG Settings
     default_embedding_model: str = "text-embedding-ada-002"
     embedding_dimension: int = 1536
+`
+        : '';
+    const ragProperties = hasRag
+        ? `
+    @property
+    def resolved_embedding_model(self) -> str:
+        """Return the embedding model with provider prefix.
+
+        If OLLAMA_BASE_URL is set and the model has no provider prefix
+        (e.g. 'nomic-embed-text' or 'nomic-embed-text:latest'),
+        automatically prepend 'ollama/' so LiteLLM/httpx can route correctly.
+        """
+        model = self.default_embedding_model
+        known_prefixes = ("ollama/", "ollama_chat/", "openai/", "anthropic/",
+                          "azure/", "cohere/", "huggingface/", "together_ai/")
+        if not any(model.startswith(p) for p in known_prefixes):
+            if self.ollama_base_url:
+                return f"ollama/{model}"
+        return model
 `
         : '';
     return `"""ChimerAI AI Service Configuration. Auto-generated by ChimerAI CLI."""
 
+import os
 from pydantic_settings import BaseSettings
 from typing import Optional
 
+# Look for .env in services/ai/ first, then fall back to project root
+_env_file = ".env" if os.path.exists(".env") else "../../.env"
+
 
 class Settings(BaseSettings):
     # Service
@@ -1772,6 +1885,9 @@ class Settings(BaseSettings):
     # API Keys (Fallback)
     openai_api_key: Optional[str] = None
 
+    # Ollama (local, no API key needed)
+    ollama_base_url: str = "http://localhost:11434"
+
     # Caching
     redis_url: str = "redis://localhost:6379"
     redis_enabled: bool = False
@@ -1779,8 +1895,23 @@ ${ragFields}
     # Default Model
     default_chat_model: str = "gpt-3.5-turbo"
 
+    @property
+    def resolved_chat_model(self) -> str:
+        """Return the chat model with provider prefix.
+
+        If OLLAMA_BASE_URL is set and the model has no provider prefix,
+        automatically prepend 'ollama/'.
+        """
+        model = self.default_chat_model
+        known_prefixes = ("ollama/", "ollama_chat/", "openai/", "anthropic/",
+                          "azure/", "cohere/", "huggingface/", "together_ai/")
+        if not any(model.startswith(p) for p in known_prefixes):
+            if self.ollama_base_url:
+                return f"ollama/{model}"
+        return model
+${ragProperties}
     class Config:
-        env_file = ".env"
+        env_file = _env_file
         extra = "ignore"
 
 

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@chimerai/cli",
-  "version": "1.2.7",
+  "version": "1.2.8",
   "description": "CLI wizard for ChimerAI starter kit — scaffold auth, RBAC, AI chat, billing and more into any Next.js project",
   "main": "./dist/index.js",
   "bin": {