feat: address PR #116 feedback — base_url, expansion_model, env rename, embed fix

Changes based on PR comments:

1. Configurable base_url for OpenAI-compatible APIs (Ollama, vLLM, Azure)
   - collections.ts: EmbeddingProviderConfig already has base_url field
   - qmd.ts: now passes base_url and expansion_model from YAML to setEmbeddingConfig
   - openai-llm.ts: constructor accepts baseURL config

2. Env var rename: QMD_OPENAI_API_KEY takes priority over OPENAI_API_KEY
   - Avoids conflict with official openai-node SDK (per @alexleach)
   - Falls back to OPENAI_API_KEY for backwards compatibility

3. generateEmbeddings bypasses LlamaCpp when using OpenAI (per @viniciushsantana)
   - OpenAI path calls API directly, no local model session needed
   - Refactored to shared runEmbedding() with pluggable embed/embedBatch fns

4. expandQuery now actually calls OpenAI for query expansion
   - Was previously returning lex-only fallback when isUsingOpenAI()
   - Now uses gpt-4o-mini via openaiLLM.expandQuery()

5. README updated with base_url, expansion_model docs

Addresses: @alexleach (env naming, base_url), @viniciushsantana (embed fix,
expansion_model, base_url YAML wiring)

Author: jonesj38

README.md

@@ -524,17 +524,20 @@ As an alternative to local embedding models, you can use OpenAI's API for faster
 embedding:
   provider: openai
   openai:
-    api_key: sk-...  # Optional, falls back to OPENAI_API_KEY env var
+    api_key: sk-...  # Optional, falls back to QMD_OPENAI_API_KEY or OPENAI_API_KEY env var
     model: text-embedding-3-small  # Optional, this is the default
+    expansion_model: gpt-4o-mini  # Optional, model for query expansion/reranking
+    base_url: https://api.openai.com/v1  # Optional, for OpenAI-compatible APIs (Ollama, vLLM, etc.)
 ```
 
 Benefits:
 - **~10x faster** than local CPU inference
 - **No GPU required** - works on any machine
 - **More reliable** - no local model loading issues
 - **Cost:** ~$0.02 per 1M tokens (very cheap)
+- **OpenAI-compatible** - works with Ollama, vLLM, Azure, and other compatible APIs via `base_url`
 
-When using OpenAI embeddings, query expansion and reranking are skipped to avoid loading local models.
+When using OpenAI embeddings, query expansion and reranking use the OpenAI API instead of local models.
 
 ## Installation
 

bun.lock

@@ -2826,6 +2826,8 @@ if (isMain) {
       openai: {
         apiKey: embeddingYamlConfig.openai?.api_key,
         embedModel: embeddingYamlConfig.openai?.model,
+        expansionModel: embeddingYamlConfig.openai?.expansion_model,
+        baseURL: embeddingYamlConfig.openai?.base_url,
       },
     });
   }

src/cli/qmd.ts

@@ -106,7 +106,7 @@ export class OpenAIEmbedding implements LLM {
 
   constructor(config: OpenAIConfig = {}) {
     this.client = new OpenAI({ 
-      apiKey: config.apiKey || process.env.OPENAI_API_KEY,
+      apiKey: config.apiKey || process.env.QMD_OPENAI_API_KEY || process.env.OPENAI_API_KEY,
       baseURL: config.baseURL,
     });
     this.embedModel = config.embedModel || DEFAULT_EMBED_MODEL;