feat: Add Ollama API compatibility layer to proxy

[mmogr]

Overview

This PR implements Ollama API compatibility for the gglib proxy, making it a drop-in replacement for Ollama. Both OpenAI (/v1/*) and Ollama (/api/*) endpoints are served simultaneously on the same port.

Changes

New Files (446 lines)

• ollama_models.rs (~520 lines): Data types, normalization, chrono-based timestamps, 10 unit tests
• ollama_stream.rs (~250 lines): SSE↔NDJSON streaming adapter using futures_util
• ollama_handlers.rs (~620 lines): All 13 Ollama route handlers with proper format translation

Modified Files

• lib.rs: Export 3 new modules
• server.rs: Merge Ollama routes, unify state struct, apply model name normalization
• Cargo.toml: Add chrono = { workspace = true }
• Cargo.lock: Updated for new dependency

API Endpoints Now Supported

✅ GET / — Root probe ("Ollama is running")
✅ GET /api/version — Version info
✅ GET /api/tags — List models
✅ POST /api/show — Model metadata
✅ GET /api/ps — Running models
✅ POST /api/chat — Chat completions (streaming + non-streaming)
✅ POST /api/generate — Text generation (streaming + non-streaming)
✅ POST /api/embed — Embeddings
✅ POST /api/embeddings — Legacy single-embedding
✅ POST /api/pull, DELETE /api/delete, POST /api/copy, POST /api/create — Stubs

Format Translation

• Client (Ollama NDJSON) → Internal (OpenAI JSON) → llama-server → Internal (OpenAI JSON) → Client (Ollama NDJSON)
• Streaming: Server-Sent Events (SSE) ↔ Newline-Delimited JSON (NDJSON)
• Model names: Strips :latest suffix automatically (e.g., phi3:latest → phi3)

Code Quality

• Single ProxyState struct eliminates state duplication
• Extracted helpers (apply_openai_options, parse_upstream_completion) reduce duplication
• 164 existing tests all pass (0 failures)
• 10 new unit tests for normalization, parsing, and timestamps

Usage

# Start proxy on Ollama's default port
gglib proxy --port 11434

# Now works with any Ollama-expecting app
curl http://localhost:11434/api/tags
curl -X POST http://localhost:11434/api/chat -d '{"model":"phi3","messages":[...]}'

Testing

cargo test --package gglib-proxy        # All 12 tests pass
cargo test --package gglib-runtime      # All 67 tests pass
cargo test --package gglib-cli          # All 34 tests pass
cargo test --package gglib-axum         # All 49 tests pass
# Total: 164 tests passing, 0 failures

Closes

#167

Breaking Changes

None. Both API surfaces coexist seamlessly without configuration."