From 03d8f49479b3446cf7f8ab9b6fdb2401584e3f12 Mon Sep 17 00:00:00 2001
From: Clawd <ai@clawd.bot>
Date: Thu, 5 Mar 2026 07:19:28 -0800
Subject: Add Ollama support as default provider, clean Embedder interface

---
 PLAN.md | 69 ++++++++++++++++++++++++++++++++++++-----------------------------
 1 file changed, 38 insertions(+), 31 deletions(-)

diff --git a/PLAN.md b/PLAN.md
index 88545b9..2c59e97 100644
--- a/PLAN.md
+++ b/PLAN.md
@@ -103,35 +103,55 @@ type Chunker interface {
 
 ## Phase 4: Embedding Generation
 
-Generate embeddings via OpenAI-compatible API (internal endpoint).
+Provider interface with Ollama (default) and OpenAI-compatible backends.
 
 **Input:** List of chunks
 **Output:** Chunks with embedding vectors
 
 ```go
+// Provider interface — easy to swap
 type Embedder interface {
     Embed(ctx context.Context, texts []string) ([][]float32, error)
+    Dimensions() int
 }
 
-type Embedder struct {
-    baseURL string  // defaults to OpenAI, configurable for internal API
+// Ollama provider (default)
+type OllamaEmbedder struct {
+    baseURL string  // default: http://localhost:11434
+    model   string  // default: nomic-embed-text
+}
+
+// OpenAI-compatible provider
+type OpenAIEmbedder struct {
+    baseURL string  // configurable for internal API
     apiKey  string
-    model   string  // "text-embedding-3-small"
+    model   string  // text-embedding-3-small, etc.
 }
 ```
 
-**Batching:** Batch chunks to minimize API calls (~100 per request).
+**Provider selection via flag:**
+```bash
+codevec index . --provider ollama --model nomic-embed-text
+codevec index . --provider openai --model text-embedding-3-small
+```
 
 **Config:**
-- `CODEVEC_API_KEY` — API key
-- `CODEVEC_BASE_URL` — Embedding API endpoint
-- `--model` flag for model selection
+- `--provider` — `ollama` (default) or `openai`
+- `--model` — model name (provider-specific defaults)
+- `CODEVEC_API_KEY` — API key (OpenAI provider)
+- `CODEVEC_BASE_URL` — Override endpoint (both providers)
+
+**Ollama models for embeddings:**
+- `nomic-embed-text` — 768 dims, good general purpose
+- `mxbai-embed-large` — 1024 dims, higher quality
 
 **Tasks:**
-- [ ] Implement OpenAI-compatible embedding client (stdlib `net/http`)
-- [ ] Support custom base URL for internal API
-- [ ] Batch requests
-- [ ] Handle rate limits with exponential backoff
+- [ ] Define `Embedder` interface
+- [ ] Implement `OllamaEmbedder` (POST to `/api/embeddings`)
+- [ ] Implement `OpenAIEmbedder` (POST to `/v1/embeddings`)
+- [ ] Provider factory based on `--provider` flag
+- [ ] Batch requests where supported
+- [ ] Handle errors gracefully (connection refused, model not found)
 
 ---
 
@@ -159,9 +179,11 @@ CREATE TABLE files (
     indexed_at INTEGER DEFAULT (unixepoch())
 );
 
+-- Dimension set at index creation based on model
+-- nomic-embed-text: 768, mxbai-embed-large: 1024, text-embedding-3-small: 1536
 CREATE VIRTUAL TABLE vec_chunks USING vec0(
     id INTEGER PRIMARY KEY,
-    embedding FLOAT[1536]
+    embedding FLOAT[768]  -- adjusted per model
 );
 ```
 
@@ -247,7 +269,9 @@ Only re-index changed files.
       "indexed_at": 1709654400
     }
   },
-  "model": "text-embedding-3-small",
+  "provider": "ollama",
+  "model": "nomic-embed-text",
+  "dimensions": 768,
   "version": 1
 }
 ```
@@ -280,7 +304,6 @@ Only re-index changed files.
 
 - TypeScript chunker (tree-sitter + TS grammar)
 - Python chunker
-- Ollama embedder for local/offline use
 - `codevec serve` HTTP API
 - Watch mode (re-index on file change)
 - Import/export index
@@ -300,22 +323,6 @@ require (
 
 ---
 
-## Estimated Effort
-
-| Phase | Effort |
-|-------|--------|
-| 1. Skeleton | 30 min |
-| 2. Walker | 1 hr |
-| 3. Chunker | 2 hr |
-| 4. Embedder | 1 hr |
-| 5. Storage | 2 hr |
-| 6. CLI | 1 hr |
-| 7. Incremental | 1 hr |
-| 8. Polish | 30 min |
-| **Total** | ~9 hr |
-
----
-
 ## Decisions
 
 1. **CLI framework:** Cobra
-- 
cgit v1.2.3