Commit graph

12 commits

Author SHA1 Message Date
Ashwin Bharambe
ea52a3ee1c minor enhancement for test fixtures 2024-11-18 22:21:17 -08:00
Ashwin Bharambe
12947ac19e
Kill "remote" providers and fix testing with a remote stack properly (#435)
# What does this PR do?

This PR kills the notion of "pure passthrough" remote providers. You
cannot specify a single provider you must specify a whole distribution
(stack) as remote.

This PR also significantly fixes / upgrades testing infrastructure so
you can now test against a remotely hosted stack server by just doing

```bash
pytest -s -v -m remote  test_agents.py \
  --inference-model=Llama3.1-8B-Instruct --safety-shield=Llama-Guard-3-1B \
  --env REMOTE_STACK_URL=http://localhost:5001
```

Also fixed `test_agents_persistence.py` (which was broken) and killed
some deprecated testing functions.

## Test Plan

All the tests.
2024-11-12 21:51:29 -08:00
Ashwin Bharambe
983d6ce2df
Remove the "ShieldType" concept (#430)
# What does this PR do?

This PR kills the notion of "ShieldType". The impetus for this is the
realization:

> Why is keyword llama-guard appearing so many times everywhere,
sometimes with hyphens, sometimes with underscores?

Now that we have a notion of "provider specific resource identifiers"
and "user specific aliases" for those and the fact that this works with
models ("Llama3.1-8B-Instruct" <> "fireworks/llama-3pv1-..."), we can
follow the same rules for Shields.

So each Safety provider can make up a notion of identifiers it has
registered. This already happens with Bedrock correctly. We just
generalize it for Llama Guard, Prompt Guard, etc.

For Llama Guard, we further simplify by just adopting the underlying
model name itself as the identifier! No confusion necessary.

While doing this, I noticed a bug in our DistributionRegistry where we
weren't scoping identifiers by type. Fixed.

## Feature/Issue validation/testing/test plan

Ran (inference, safety, memory, agents) tests with ollama and fireworks
providers.
2024-11-12 12:37:24 -08:00
Ashwin Bharambe
09269e2a44
Enable sane naming of registered objects with defaults (#429)
# What does this PR do? 

This is a follow-up to #425. That PR allows for specifying models in the
registry, but each entry needs to look like:

```yaml
- identifier: ...
  provider_id: ...
  provider_resource_identifier: ...
```

This is headache-inducing.

The current PR makes this situation better by adopting the shape of our
APIs. Namely, we need the user to only specify `model-id`. The rest
should be optional and figured out by the Stack. You can always override
it.

Here's what example `ollama` "full stack" registry looks like (we still
need to kill or simplify shield_type crap):
```yaml
models:
- model_id: Llama3.2-3B-Instruct
- model_id: Llama-Guard-3-1B
shields:
- shield_id: llama_guard
  shield_type: llama_guard
```

## Test Plan

See test plan for #425. Re-ran it.
2024-11-12 11:18:05 -08:00
Ashwin Bharambe
d9d271a684
Allow specifying resources in StackRunConfig (#425)
# What does this PR do? 

This PR brings back the facility to not force registration of resources
onto the user. This is not just annoying but actually not feasible
sometimes. For example, you may have a Stack which boots up with private
providers for inference for models A and B. There is no way for the user
to actually know which model is being served by these providers now (to
be able to register it.)

How will this avoid the users needing to do registration? In a follow-up
diff, I will make sure I update the sample run.yaml files so they list
the models served by the distributions explicitly. So when users do
`llama stack build --template <...>` and run it, their distributions
come up with the right set of models they expect.

For self-hosted distributions, it also allows us to have a place to
explicit list the models that need to be served to make the "complete"
stack (including safety, e.g.)

## Test Plan

Started ollama locally with two lightweight models: Llama3.2-3B-Instruct
and Llama-Guard-3-1B.

Updated all the tests including agents. Here's the tests I ran so far:

```bash
pytest -s -v -m "fireworks and llama_3b" test_text_inference.py::TestInference \
  --env FIREWORKS_API_KEY=...

pytest -s -v -m "ollama and llama_3b" test_text_inference.py::TestInference 

pytest -s -v -m ollama test_safety.py

pytest -s -v -m faiss test_memory.py

pytest -s -v -m ollama  test_agents.py \
  --inference-model=Llama3.2-3B-Instruct --safety-model=Llama-Guard-3-1B
```

Found a few bugs here and there pre-existing that these test runs fixed.
2024-11-12 10:58:49 -08:00
Ashwin Bharambe
c1f7ba3aed
Split safety into (llama-guard, prompt-guard, code-scanner) (#400)
Splits the meta-reference safety implementation into three distinct providers:

- inline::llama-guard
- inline::prompt-guard
- inline::code-scanner

Note that this PR is a backward incompatible change to the llama stack server. I have added deprecation_error field to ProviderSpec -- the server reads it and immediately barfs. This is used to direct the user with a specific message on what action to perform. An automagical "config upgrade" is a bit too much work to implement right now :/

(Note that we will be gradually prefixing all inline providers with inline:: -- I am only doing this for this set of new providers because otherwise existing configuration files will break even more badly.)
2024-11-11 09:29:18 -08:00
Dinesh Yeduguru
d800a16acd
Resource oriented design for shields (#399)
* init

* working bedrock tests

* bedrock test for inference fixes

* use env vars for bedrock guardrail vars

* add register in meta reference

* use correct shield impl in meta ref

* dont add together fixture

* right naming

* minor updates

* improved registration flow

* address feedback

---------

Co-authored-by: Dinesh Yeduguru <dineshyv@fb.com>
2024-11-08 12:16:11 -08:00
Ashwin Bharambe
694c142b89
Add provider deprecation support; change directory structure (#397)
* Add provider deprecation support; change directory structure

* fix a couple dangling imports

* move the meta_reference safety dir also
2024-11-07 13:04:53 -08:00
Ashwin Bharambe
064d2a5287
Remove the safety adapter for Together; we can just use "meta-reference" (#387) 2024-11-06 17:36:57 -08:00
Ashwin Bharambe
994732e2e0
impls -> inline, adapters -> remote (#381) 2024-11-06 14:54:05 -08:00
Ashwin Bharambe
7cf4c905f3 add support for remote providers in tests 2024-11-04 20:30:46 -08:00
Ashwin Bharambe
ffedb81c11
Significantly simpler and malleable test setup (#360)
* Significantly simpler and malleable test setup

* convert memory tests

* refactor fixtures and add support for composable fixtures

* Fix memory to use the newer fixture organization

* Get agents tests working

* Safety tests work

* yet another refactor to make this more general

now it accepts --inference-model, --safety-model options also

* get multiple providers working for meta-reference (for inference + safety)

* Add README.md

---------

Co-authored-by: Ashwin Bharambe <ashwin@meta.com>
2024-11-04 17:36:43 -08:00