Skip to content

Sync with upstream develop#23

Merged
m-misiura merged 9 commits into
trustyai-explainability:developfrom
m-misiura:sync-with-upstream-develop
Feb 17, 2026
Merged

Sync with upstream develop#23
m-misiura merged 9 commits into
trustyai-explainability:developfrom
m-misiura:sync-with-upstream-develop

Conversation

@m-misiura
Copy link
Copy Markdown
Collaborator

Need to sync with upstream develop since resolution to this issue is required

Pouyanpi and others added 8 commits February 10, 2026 16:06
…sk-specific LLM support (NVIDIA-NeMo#1603)

Fixes a naming mismatch bug where the action `generate_next_step`
(singular) didn't match the Task enum value `generate_next_steps`
(plural). This mismatch prevented task-specific LLM configuration from
working correctly.

When users configured a task-specific LLM with `type:
generate_next_steps` in their config, the runtime would look for
`generate_next_step_llm` (based on the action name) but the config
registered `generate_next_steps_llm`, causing the task to fall back to
the main LLM instead.
* docs(llmrails): add check_async/check reference page

---------

Co-authored-by: Miyoung Choi <miyoungc@nvidia.com>
* docs(integration): add Agent Middleware documentation
---------

Co-authored-by: Miyoung Choi <miyoungc@nvidia.com>
…capture (NVIDIA-NeMo#1624)

* refactor(streaming)!: remove stream_usage and fix streaming metadata capture

* Update docs/run-rails/using-python-apis/streaming.md

Signed-off-by: Pouyan <13303554+Pouyanpi@users.noreply.github.com>
…on (NVIDIA-NeMo#1619)

* Update Fiddler Guardrails API to match new specification

- Change Safety API from 'prompt' to 'input' field
- Change Faithfulness API to use strings instead of lists
- Add fdl_roleplaying category to safety detection
- Add unit tests to verify request payload format
…eMo#1610)

Update repository URL in Pangea integration User-Agent header.
…egration (NVIDIA-NeMo#1578) (NVIDIA-NeMo#1611)

* fix(guardrails_ai): add valid alias to action results for Colang flows (NVIDIA-NeMo#1578)

Fixes NVIDIA-NeMo#1578

GuardrailsAI actions previously returned only {"validation_result": ...},
while the GuardrailsAI flows in Colang expect $result["valid"]` to decide
whether to block. This caused a KeyError: 'valid' at runtime.

This change updates validate_guardrails_ai_input and
validate_guardrails_ai_output to return a dict containing both
validation_result and a valid boolean derived via
guardrails_ai_validation_mapping. Colang flows can keep using
$result["valid"], and existing callers of validation_result remain
backwards compatible.
* pysdk reference autogeneration

* fix the lock file

* fix build
@m-misiura m-misiura force-pushed the sync-with-upstream-develop branch from 1499cbe to 0211ecc Compare February 16, 2026 13:46
@m-misiura m-misiura force-pushed the sync-with-upstream-develop branch from e381199 to 2bb486a Compare February 16, 2026 13:57
@m-misiura m-misiura merged commit fb42a54 into trustyai-explainability:develop Feb 17, 2026
3 of 12 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

5 participants