**Remove stale model entries and tidy configuration**

Add qwen3.5‑397b‑a17b model entries (NVIDIA and Ollama) and replace all hard‑coded NVIDIA `api_base` URLs with `os.environ/NVIDIA_BASE_URL` in `litellm_config.yaml`.

Add `order`, `rpm`, `max_parallel_requests`, and `tags` fields to model configurations; extend `context_window_fallbacks` and `fallbacks` for planner, executor, and leaf‑operator.

Add `power` NVIDIA‑only alias with ordered frontier ladder, configure its models in `litellm_config.yaml`, and tighten retry limits (set `allowed_fails` to 1 and reduce timeout/rate‑limit retries to 1).

Add `input_type: "passage"` to NVIDIA nv‑embedcode‑7b‑v1 model entries in litellm_config.yaml.

Add `encoding_format: "float"` to NVIDIA nv‑embedcode‑7b‑v1 model entries in litellm_config.yaml.

Add `encoding_format: "float"` to NVIDIA nv‑embedcode‑7b‑v1 model entries in litellm_config.yaml.

Add code embedding model (nvidia/nv-embedcode-7b-v1) configuration to litellm_config.yaml.

Add NVIDIA nv‑embedcode‑7b‑v1 embedding model configuration to litellm_config.yaml.

Add Gemini rate‑limits markdown, Gemini models JSON, and NVIDIA model catalog markdown files.

Add extensive model configurations (chat, reasoning, coding, multimodal, size bands, etc.) to litellm_config.yaml.

```markdown

Add `database_url` to `litellm_config.yaml` and document admin panel requirements in README.

initial commit