You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Requesting a new Haystack integration for Synthorai (https://synthorai.io), an OpenAI/Anthropic-compatible LLM gateway routing to 122 models across 11 upstream providers (Claude, GPT, Gemini, GLM, Kimi, DeepSeek, Qwen, etc.) at direct upstream pricing, no markup. Docs: https://synthorai.io/docs. Use case: one API key/base URL for both Claude-family and GPT-family models plus a longer tail of Chinese-origin models, without juggling multiple provider SDKs in a Haystack pipeline.
Adoption signals
Being upfront that Synthorai is early-stage, so these numbers are modest right now — flagging this as a heads-up request rather than expecting immediate prioritization, in case it's useful to have on record:
GitHub stars: N/A (Synthorai itself isn't an open-source project; it's a hosted gateway)
PyPI downloads: N/A, no Python SDK published yet — plain HTTP/OpenAI-SDK-compatible
Release activity: Live production service, continuously updated
Maintenance: Actively developed and maintained
Haystack community demand: None yet that I'm aware of — this is a proactive request, not in response to existing community asks
Following the OpenRouter integration (#1723) as a template: a new integrations/synthorai/ package with a SynthoraiChatGenerator wrapping the existing OpenAI-compatible generator machinery, defaulting api_base_url to https://synthorai.io/v1 and reading SYNTHORAI_API_KEY. Plain OpenAI-compatible surface, no custom request/response transformation needed.
I'm not attempting the PR myself yet given the adoption-signal bar described above — happy to if the team thinks it's worth building despite the early-stage numbers, and can use the scaffolding script per CONTRIBUTING.md at that point.
Summary and motivation
Requesting a new Haystack integration for Synthorai (https://synthorai.io), an OpenAI/Anthropic-compatible LLM gateway routing to 122 models across 11 upstream providers (Claude, GPT, Gemini, GLM, Kimi, DeepSeek, Qwen, etc.) at direct upstream pricing, no markup. Docs: https://synthorai.io/docs. Use case: one API key/base URL for both Claude-family and GPT-family models plus a longer tail of Chinese-origin models, without juggling multiple provider SDKs in a Haystack pipeline.
Adoption signals
Being upfront that Synthorai is early-stage, so these numbers are modest right now — flagging this as a heads-up request rather than expecting immediate prioritization, in case it's useful to have on record:
Detailed design
Following the OpenRouter integration (#1723) as a template: a new
integrations/synthorai/package with aSynthoraiChatGeneratorwrapping the existing OpenAI-compatible generator machinery, defaultingapi_base_urltohttps://synthorai.io/v1and readingSYNTHORAI_API_KEY. Plain OpenAI-compatible surface, no custom request/response transformation needed.I'm not attempting the PR myself yet given the adoption-signal bar described above — happy to if the team thinks it's worth building despite the early-stage numbers, and can use the scaffolding script per CONTRIBUTING.md at that point.