SyGra-v2.0.0

A feature-rich release focused on UI based flow building and generation, expanded provider support (Vertex AI/Bedrock), multimodal additions (audio transcription, TTS, image generation), ServiceNow data source/sink integration, multi dataset support as source or sink, observability/metrics, enhanced tooling (publish workflow, pre-commit, CI fixes), data quality improvements (self-refinement and semantic deduping) and multiple bug fixes (lambda/node stability, structured output/performance, tool-call fixes).

Major new features & enhancements

1. SyGra Studio — SyGra Studio is a visual workflow builder that replaces manual YAML editing with an interactive drag-and-drop interface. It also allows you to execute a task, monitor during execution and view the result along with metadata like latency, token usage etc.
2. Tools support in LLM Node — Added first-class support for using tools inside the LLM Node, enabling richer tool-based workflows. This helps generating training or eval data for agentic flow. Agent node already support tool execution, however this enables generating tool calls for training or evaluation.
3. Multimodal output support
   • Audio transcription support — New transcription capability for audio inputs.
   • OpenAI TTS models support — Added TTS model support enabling text → speech flows.
   • OpenAI image generation support — Added image generation via OpenAI (image) endpoints.
4. Provider / model integrations expanded — LiteLLM is now the default backed for all model types enabling smoother integration of new model types. Added support for Vertex AI, Bedrock providers and gpt-4o-audio model.
5. ServiceNow Instance integration — ServiceNow as data source or sink(useful for enterprise data pipelines).
6. Multiple Datasets - Added support for multiple datasets as data source/sink, allowing users to use any number/mix of datasets for reading or pushing data.
7. Metadata & observability improvements — Added metadata tracking (latency percentiles, node costs) and aggregator metrics for platform observability.
8. Evaluation Metrics — Metrics added to be used for evaluating model response or writing evaluation flow.
9. Semantic Deduping — Added this feature to dedup the generated synthetic data.
10. Self-Refinement recipe — Generates datasets that include judge scores and a reflection trajectory for each prompt.

Notable fixes & stability improvements

1. Fixes for structured output performance and null tool-call errors that improve runtime reliability.
2. Lambda node failure fixes — addressed issues causing lambda node failures.
3. CI / test fixes — resolved CI cache and async test-case failures for smoother CI runs.

Tooling, packaging & development workflow

1. Publish workflow added to the repo for automated publishing.
2. Migration from Poetry to uv packaging (packaging/environment changes).
3. Pre-commit config and other developer-experience improvements.
4. uv - Making the package manager simpler and faster.

What's Changed

• Update README.md by @vipul-mittal in #46
• [Enhancement] Add publish workflow by @psriramsnc in #47
• [Fix] self.id_column Type Mismatch in BaseTaskExecutor by @thejishnunair in #45
• Support for openai endpoints by @amitsnow in #48
• [Enhancement] Updated pyproject with fixed lib versions by @psriramsnc in #50
• [Enhancement] Refactor graph configuration management and custom config support by @zephyrzilla in #53
• [Fix] Multi Auth Token issue by @psriramsnc in #51
• [Enhancement] Add support for Chat Template params by @psriramsnc in #52
• [Fix] Fix ci cache issue by @psriramsnc in #56
• [Enhancement] Add pre-commit config by @zephyrzilla in #55
• [Enhancement] Adding support for Openai TTS models by @amitsnow in #54
• [Fix] Fix Async test cases by @psriramsnc in #57
• OpenAI image generation support by @amitsnow in #58
• [Enhancement] Support tools in LLM Node by @psriramsnc in #60
• [Enhancement] Add metadata tracking system with latency percentiles and node costs by @zephyrzilla in #59
• [Enhancement] Adding support for gpt-4o-audio by @amitsnow in #61
• [Enhancement] LiteLLM Integration by @psriramsnc in #62
• [Enhancement] Support Vertex AI, Bedrock Models by @psriramsnc in #64
• [Fix] Fix lambda node failures by @psriramsnc in #67
• [Enhancement] Add ServiceNow Data Handler (Source/Sink) Integration by @zephyrzilla in #65
• [Fix] Fix Structured output Performance issues by @psriramsnc in #69
• [Enhancement/Document] ServiceNow instance support document by @bidyapati-p in #68
• [Fix] Fix for null tool call error by @Sidharthenee333 in #70
• [Enhancement] Audio Transcription Support by @amitsnow in #63
• [Enhancement] Migrate project environment & packaging from Poetry to uv by @psriramsnc in #71
• [Enhancement] Adding aggregator metrics to platform for generic or task specific usage by @omkar-anustoop-ai in #66
• [Fix] Fix tools calls issue with openai models by @psriramsnc in #73
• [Fix] Update README.md by @vipul-mittal in #74
• Doc Update and fixes by @amitsnow in #75
• [Enhancement] Support for multiple dataset as source, sink and in prompts by @bidyapati-p in #72
• [Enhancement] Adding pass@k and pass^k metrics by @NiraliPopat in #78
• [Enhancement] Adding unit metrics to platform for generic or task specific usage by @omkar-anustoop-ai in #79
• [Enhancement] Add support for custom proxy models by @Sidharthenee333 in #101
• [Enhancement] SyGra Studio - UI Workflow Builder and Dashboard by @zephyrzilla in #102
• Sygra Studio bug fixes by @bidyapati-p in #116
• [Enhancement] Add Self Refinement Feature by @psriramsnc in #105
• [Enhancement] Add Support for Semantic Deduplication by @psriramsnc in #104
• [Enhancement] Studio UI refresh and documentation by @zephyrzilla in #120
• SyGra studio minor issues by @bidyapati-p in #118
• [Fix] Framework and Studio fixes by @zephyrzilla in #125
• Support for sync and async lambda function by @bidyapati-p in #122

New Contributors

• @thejishnunair made their first contribution in #45
• @Sidharthenee333 made their first contribution in #70

Full Changelog: v1.0.0...v2.0.0