Initial commit: llm-router — smart OpenAI/Anthropic/Ollama router

A stdlib pre-router in front of Ollama with LiteLLM backend:
- auto model selection by content/tools/modality, with fallbacks
- OpenAI /v1, Anthropic /v1/messages, and Ollama-native /api/* endpoints
- Whisper-shaped /v1/audio/transcriptions + in-chat audio
- key-based fleet policies (e.g. force a client onto uncensored models)
- optional Bearer auth; launchd/systemd service install
- benchmark harnesses (speed, quality, agentic tool use) with sample results

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

scripts/pull-models.sh

@@ -0,0 +1,39 @@
+#!/usr/bin/env bash
+# Pull a model fleet into Ollama. EDIT the MODELS list to taste, then run.
+# Continues past failures; prints a summary. Ollama resumes partial downloads.
+#
+# The names here must line up with the routing map in router.py (M / UNCENSORED).
+# If you change models, update router.py accordingly (see docs/ROUTING.md).
+set -u
+
+MODELS=(
+  # --- general / small ---
+  "qwen3:8b"
+  "qwen3:14b"
+  # --- reasoning / agentic ---
+  "qwen3.6:35b-a3b"
+  "gpt-oss:20b"
+  "glm-4.7-flash"
+  # --- coding ---
+  "qwen3-coder:30b"
+  # --- vision / OCR ---
+  "qwen3-vl:8b"
+  "qwen3-vl:30b-a3b-instruct"
+  # --- multimodal (vision + audio) ---
+  "gemma4:e4b"          # ROUTER_AUDIO_MODEL default — needed for /v1/audio/*
+  "gemma4:12b"
+  "gemma4:26b"
+
+  # --- example: a GGUF straight from Hugging Face (see docs/ROUTING.md) ---
+  # "hf.co/USER/REPO:Q4_K_M"
+)
+
+ok=(); fail=(); i=0; total=${#MODELS[@]}
+for m in "${MODELS[@]}"; do
+  i=$((i+1))
+  echo "=== [$i/$total] pulling $m ==="
+  if ollama pull "$m"; then ok+=("$m"); else echo "!! failed: $m"; fail+=("$m"); fi
+done
+
+echo ""; echo "done. ${#ok[@]}/$total succeeded."
+[ ${#fail[@]} -gt 0 ] && { echo "failed:"; printf '  - %s\n' "${fail[@]}"; }