GitLab

Publish `generation_config.json` from worker to ingress, as part of Model Deployment Card. That allows ingress to read key fields out of it. Gemma 3 4B+ has some important information that's only in there.

Example:
```
dynamo-run out=<engine> <model> --kv-cache-block-size 64
```

In a distributed system this goes on the worker node and is propagated to ingress via the model deployment card.

Previously hard coded to 16, which is now the default.

- Load context_length from model. Closes #1172
- Store context length and KV cache block size in Model Deployment Card #1170

Removed the hard coded sleeps, explained what we're testing.

Closes https://github.com/ai-dynamo/dynamo/issues/1132

The race condition is that `apply_event` sends a message on a channel, it does not directly apply the event. At some later point the tokio runtime schedules the task running the channel receiver, which applies the event. If that had not happened yet the test would fail.

- Stop advertising a model when it's last instance stops. Previously was when any instance stops.
- Faster locks on model manager.
- Move discovery code out of http, as it is used by all inputs.

Implements OpenAI embeddings (interface only).

- Adds ModelType::Embedding
- Adds OpenAI embedding request/response structs
- Adds support for embedding model discovery

. New mistralrs and llamacpp version
. mistralrs: Handle Gemma 3 and Llama 4 as vision models
. Update the dynamo-run docs to use Qwen 3
. Our pre-processor now supports Llama 4's newer multi-modal `config.json`
. Upgrade minijinja to handle Qwen 3's prompt template

For Llama 4 we'll need to limit the max seq len. vllm says:
> To serve at least one request with the models's max seq len (10485760), (240.00 GiB KV cache is needed,...

I was able to run Llama 4 with llamacpp and a quantized GGUF, with Dynamo doing the pre-processing.

Adding this to a Python script makes it register on the network so that `dynamo-run` can discover it and send it requests:
```
from dynamo.llm import register_llm

MODEL = "Qwen/Qwen2.5-0.5B-Instruct"
await register_llm(endpoint, MODEL, 3)
```

Full vllm example, with pre-processing in dynamo:
- `dynamo-run in=text out=dyn://dynamo.backend.generate`
- `cd lib/bindings/python/examples/hello_world`
- `python server_vllm.py`

This builds on top of the work to move pre-processor to ingress side. It means we can decouple Rust and Python using NATS as the bus.

The `register_llm` call does this:

- Download the model from HF if necessary
- Load the model deployment card from the HF folder or extract from GGUF
- Push the tokenizer config etc into NATS object store so ingress can access it from a different machine
- Publish the model deployment card to ETCD

This will allow an ingress-side pre-processor to see it without needing a model checkout.

Currently pre-processing is done in the worker, which has access to the model deployment card ("MDC") files (`config.json`, `tokenizer.json` and `tokenizer_config.json`) locally. We want to move the pre-processor to the ingress side to support KV routing. That requires ingress side (i.e the HTTP server), on a different machine than the worker to be able to see those three files.

To support that this PR makes the worker upload the contents of those files to the NATS object store, and publishes the MDC with those NATS urls to the key-value store. 

The key-value store has an interface so any store (nats, etcd, redis, etc) can be supported. Implementations for memory and NATS are provided.

Fetching the MDC from the store, doing pre-processing ingress side, and publishing a card backed by a GGUF, are all for a later commit.

Part of #743 

This lets us do:
```
dynamo-run out=llamacpp <gguf_file>
```

Previously a `--model-config <hf-repo>` was also required, to configure our tokenizer.

Co-authored-by: Biswa Panda <biswa.panda@gmail.com>

Co-authored-by: Graham King <grahamk@nvidia.com>

Co-authored-by: Graham King <grahamk@nvidia.com>

Add backend type `EngineConfig::StaticCore` that wraps the engine in a preprocessor (prompt templating and tokenization).

Add example engine `echo_core` (`out=echo_core`) which takes and returns tokens. A nice side effect is that it echos the full prompt template with system prompt, whereas `echo_full` echos only user prompt.

![image](https://github.com/user-attachments/assets/27ec0a7b-a27d-4e69-96ea-1ffa0822ea90)

Signed-off-by: Ryan McCormick <rmccormick@nvidia.com>

Signed-off-by: Neelay Shah <neelays@nvidia.com>
Co-authored-by: Ryan McCormick <rmccormick@nvidia.com>