Switch from embeddinggemma:300m to qwen3-embedding:0.6b

CLAUDE.md

@@ -77,7 +77,7 @@ pnpm test
 | Container Runtime | Docker + Docker Compose |
 | Cloud LLM | Claude API (claude-opus-4-5-20251101) |
 | Local LLM | Ollama (glm-4.7:cloud for HyDE/RAG) |
-| Embeddings | embeddinggemma:300m (768 dim) |
+| Embeddings | qwen3-embedding:0.6b (1024 dim) |
 | Frontend | React / Next.js (future) |
 
 ## Multi-Agent Workspace Structure
@@ -359,7 +359,7 @@ ROBOCO_RAG_USE_HYDE=true
 ROBOCO_RAG_USE_HYBRID_SEARCH=true
 
 # AI/LLM
-ROBOCO_DEFAULT_EMBEDDING_MODEL=embeddinggemma:300m
+ROBOCO_DEFAULT_EMBEDDING_MODEL=qwen3-embedding:0.6b
 ROBOCO_LOCAL_LLM_MODEL=glm-4.7:cloud
 ROBOCO_LOCAL_LLM_BASE_URL=http://roboco-ollama:11434/v1
 ROBOCO_OLLAMA_BASE_URL=http://roboco-ollama:11434
@@ -387,7 +387,7 @@ The startup order is critical due to dependencies:
 postgres ──┐
 redis ─────┼──> ollama ──> ollama-init ──> orchestrator
            │        │            │
-           │        │            └── Pulls embeddinggemma:300m, glm-4.7:cloud
+           │        │            └── Pulls qwen3-embedding:0.6b, glm-4.7:cloud
            │        └── Healthcheck: ollama list
            └── Healthcheck: pg_isready, redis-cli ping
 ```
@@ -404,7 +404,7 @@ Ollama provides two APIs:
 - `/v1/*` - OpenAI-compatible API (for LLM chat/completion)
 - `/api/*` - Native Ollama API (for embeddings, model management)
 
-The embedder uses `/api/embed` endpoint with the `embeddinggemma:300m` model.
+The embedder uses `/api/embed` endpoint with the `qwen3-embedding:0.6b` model.
 
 **Environment variables for Docker:**
 ```bash

README.md

@@ -176,7 +176,7 @@ uv run mypy roboco/
 | Vector Store | pgvector (via piragi) |
 | Cache/Queue | Redis |
 | RAG Library | piragi |
-| Embeddings | embeddinggemma:300m (sentence-transformers) |
+| Embeddings | qwen3-embedding:0.6b (sentence-transformers) |
 | Local LLM | Ollama (glm-4.7:cloud) |
 | Cloud LLM | Claude API (Anthropic) |
 | Package Manager | uv |

docker-compose.yaml

@@ -72,10 +72,10 @@ services:
     command:
       - |
         set -e
-        echo "=== Pulling embedding model (embeddinggemma:300m) ==="
+        echo "=== Pulling embedding model (qwen3-embedding:0.6b) ==="
         # Ollama /api/pull streams JSON lines until complete - consume full stream
         # Note: $$ escapes $ for docker-compose variable substitution
-        curl -sN http://ollama:11434/api/pull -d '{"name":"embeddinggemma:300m"}' | while read -r line; do
+        curl -sN http://ollama:11434/api/pull -d '{"name":"qwen3-embedding:0.6b"}' | while read -r line; do
           status=$$(echo "$$line" | grep -o '"status":"[^"]*"' | cut -d'"' -f4)
           [ -n "$$status" ] && echo "  $$status"
         done
@@ -85,7 +85,7 @@ services:
           [ -n "$$status" ] && echo "  $$status"
         done
         echo "=== Verifying models are available ==="
-        curl -sf http://ollama:11434/api/tags | grep -q "embeddinggemma" && echo "  embeddinggemma: OK"
+        curl -sf http://ollama:11434/api/tags | grep -q "qwen3-embedding" && echo "  qwen3-embedding: OK"
         curl -sf http://ollama:11434/api/tags | grep -q "glm-4.7" && echo "  glm-4.7: OK"
         echo "=== All models ready! ==="
 
@@ -221,7 +221,7 @@ services:
       # Ollama (use container name)
       ROBOCO_LOCAL_LLM_BASE_URL: http://roboco-ollama:11434/v1
       ROBOCO_LOCAL_LLM_MODEL: glm-4.7:cloud
-      ROBOCO_DEFAULT_EMBEDDING_MODEL: embeddinggemma:300m
+      ROBOCO_DEFAULT_EMBEDDING_MODEL: qwen3-embedding:0.6b
       ROBOCO_OLLAMA_BASE_URL: http://roboco-ollama:11434
       # Host paths for spawning agent containers (required for Docker-in-Docker)
       # IMPORTANT: These must be ABSOLUTE paths on the host filesystem

docker-compose.yml

@@ -72,10 +72,10 @@ services:
     command:
       - |
         set -e
-        echo "=== Pulling embedding model (embeddinggemma:300m) ==="
+        echo "=== Pulling embedding model (qwen3-embedding:0.6b) ==="
         # Ollama /api/pull streams JSON lines until complete - consume full stream
         # Note: $$ escapes $ for docker-compose variable substitution
-        curl -sN http://ollama:11434/api/pull -d '{"name":"embeddinggemma:300m"}' | while read -r line; do
+        curl -sN http://ollama:11434/api/pull -d '{"name":"qwen3-embedding:0.6b"}' | while read -r line; do
           status=$$(echo "$$line" | grep -o '"status":"[^"]*"' | cut -d'"' -f4)
           [ -n "$$status" ] && echo "  $$status"
         done
@@ -85,7 +85,7 @@ services:
           [ -n "$$status" ] && echo "  $$status"
         done
         echo "=== Verifying models are available ==="
-        curl -sf http://ollama:11434/api/tags | grep -q "embeddinggemma" && echo "  embeddinggemma: OK"
+        curl -sf http://ollama:11434/api/tags | grep -q "qwen3-embedding" && echo "  qwen3-embedding: OK"
         curl -sf http://ollama:11434/api/tags | grep -q "glm-4.7" && echo "  glm-4.7: OK"
         echo "=== All models ready! ==="
 
@@ -221,7 +221,7 @@ services:
       # Ollama (use container name)
       ROBOCO_LOCAL_LLM_BASE_URL: http://roboco-ollama:11434/v1
       ROBOCO_LOCAL_LLM_MODEL: glm-4.7:cloud
-      ROBOCO_DEFAULT_EMBEDDING_MODEL: embeddinggemma:300m
+      ROBOCO_DEFAULT_EMBEDDING_MODEL: qwen3-embedding:0.6b
       ROBOCO_OLLAMA_BASE_URL: http://roboco-ollama:11434
       # Host paths for spawning agent containers (required for Docker-in-Docker)
       # IMPORTANT: These must be ABSOLUTE paths on the host filesystem

docs/rag/architecture/config-reference.md

@@ -41,8 +41,8 @@ Environment variables for RoboCo (prefix: `ROBOCO_`).
 
 | Variable | Default | Description |
 |----------|---------|-------------|
-| `ROBOCO_DEFAULT_EMBEDDING_MODEL` | `embeddinggemma:300m` | Embedding model |
-| `ROBOCO_EMBEDDING_DIMENSIONS` | `768` | Embedding dimensions |
+| `ROBOCO_DEFAULT_EMBEDDING_MODEL` | `qwen3-embedding:0.6b` | Embedding model |
+| `ROBOCO_EMBEDDING_DIMENSIONS` | `1024` | Embedding dimensions |
 | `ROBOCO_RAG_CHUNK_STRATEGY` | `fixed` | fixed/semantic/hierarchical/contextual |
 | `ROBOCO_RAG_CHUNK_SIZE` | `512` | Base chunk size |
 | `ROBOCO_RAG_CHUNK_SIZE_DOCS` | `1536` | Chunk size for docs |

roboco/config.py

@@ -168,12 +168,12 @@ class Settings(BaseSettings):
 
     # Default models
     default_embedding_model: str = Field(
-        default="embeddinggemma:300m",
-        description="Embedding model (Ollama). embeddinggemma:300m for quality.",
+        default="qwen3-embedding:0.6b",
+        description="Embedding model. Qwen3 Embedding for quality + 32K context.",
     )
     embedding_dimensions: int = Field(
-        default=768,
-        description="Embedding dimensions (768 for embeddinggemma/nomic-embed-text)",
+        default=1024,
+        description="Embedding dimensions (1024 for qwen3-embedding)",
     )
 
     # Local LLM for RAG (HyDE, reranking, etc.)

roboco/services/optimal.py

@@ -51,7 +51,7 @@ from roboco.services.optimal_brain.indexes.reviews import (
 logger = structlog.get_logger()
 
 # Max chars per citation content - increased for better synthesis quality
-# embeddinggemma:300m retrieves higher quality chunks, so more context helps
+# qwen3-embedding:0.6b retrieves higher quality chunks, so more context helps
 MAX_CONTENT_CHARS = 800
 
 

roboco/services/optimal_brain/indexes/base.py

@@ -33,7 +33,7 @@ class IndexConfig:
     use_hyde: bool = True
     use_hybrid_search: bool = True
     use_cross_encoder: bool = False
-    embedding_model: str = "embeddinggemma:300m"
+    embedding_model: str = "qwen3-embedding:0.6b"
     llm_model: str = "glm-4.7:cloud"
     llm_base_url: str = "http://roboco-ollama:11434/v1"
 

roboco/services/optimal_brain/ollama_embedder.py

@@ -145,7 +145,7 @@ class OllamaEmbedder:
 
     def __init__(
         self,
-        model: str = "embeddinggemma:300m",
+        model: str = "qwen3-embedding:0.6b",
         base_url: str | None = None,
         timeout: float = 120.0,
         max_concurrent: int = MAX_CONCURRENT_BATCHES,

roboco/services/optimal_brain/shared_embedder.py

@@ -2,7 +2,7 @@
 Shared Embedder Singleton
 
 Provides a single embedder instance shared across all index plugins.
-Supports both Ollama models (embeddinggemma, etc.) and SentenceTransformers (BGE, etc.).
+Supports both Ollama models (qwen3-embedding, ...) and SentenceTransformers (BGE, ...).
 """
 
 import asyncio
@@ -45,6 +45,7 @@ class Embedder(Protocol):
 
 # Known Ollama embedding models
 OLLAMA_EMBEDDING_MODELS = {
+    "qwen3-embedding",
     "embeddinggemma",
     "nomic-embed-text",
     "mxbai-embed-large",