From 6710ec328ede3e47d92b3ee0c614c32a9834e541 Mon Sep 17 00:00:00 2001 From: DABH Date: Mon, 26 Oct 2026 19:42:57 -0500 Subject: [PATCH] Correct retry-history and dedicated-endpoint guidance on the open-weight provider pages Follow-up to review comments on #5574. Event History records an Activity's result, attempt count, and last failure, not every attempt, so the retry paragraph now says so and points at Worker logs and the pending Activity's last failure for individual attempts. The Baseten and Together AI pages now say explicitly that the sample's provider prefix only reaches the shared endpoint, and that a dedicated deployment needs its own PROVIDERS entry with a distinct name and base URL. --- docs/develop/python/integrations/baseten.mdx | 8 ++++++-- docs/develop/python/integrations/fireworks-ai.mdx | 5 +++-- docs/develop/python/integrations/together-ai.mdx | 10 ++++++---- 3 files changed, 15 insertions(+), 8 deletions(-) diff --git a/docs/develop/python/integrations/baseten.mdx b/docs/develop/python/integrations/baseten.mdx index d126e99d91..c4fc0a1fd3 100644 --- a/docs/develop/python/integrations/baseten.mdx +++ b/docs/develop/python/integrations/baseten.mdx @@ -70,7 +70,9 @@ license that Baseten serves on Model APIs, where tool calling is supported for e [model library page](https://www.baseten.co/library/gpt-oss-120b/) and `GET https://inference.baseten.co/v1/models` describe the catalog. A [dedicated deployment](https://docs.baseten.co/inference/calling-your-model) has its own base URL, `https://model-.api.baseten.co/environments/production/sync/v1`, takes the same API key, and expects the model -name that the deployment serves, so in the sample it is a second entry in `PROVIDERS`. +name that the deployment serves. The sample's `baseten/` prefix only reaches Model APIs: to call a deployment, add a +`PROVIDERS` entry with a distinct name and that base URL, such as `baseten-dedicated`, and use its prefix in the model +name. A `ModelProvider` routes a `/` name to the matching endpoint and returns the Chat Completions model that the Agents SDK calls: @@ -252,7 +254,9 @@ the same Worker. The plugin classifies Baseten's HTTP errors. Status 408, 409, 429, and any 5xx retry under the Activity's Retry Policy, and a `Retry-After` header becomes the next retry delay. Connection errors and client-side timeouts retry as well. Other 4xx responses, such as an invalid API key or an unknown model name, fail the Activity without retrying, as does a model -name whose provider has no key on the Worker. Each attempt, its failure, and the final result are in Event History. +name whose provider has no key on the Worker. Event History records the Activity's result, its attempt count, and its +last failure, not every attempt. Individual failed attempts appear in the Worker's logs and, while the Activity is still +retrying, as the pending Activity's last failure when you describe the Workflow Execution. Baseten enforces requests-per-minute and tokens-per-minute limits that depend on your [account tier](https://docs.baseten.co/inference/model-apis/pricing-and-limits) and answers `429` when either is diff --git a/docs/develop/python/integrations/fireworks-ai.mdx b/docs/develop/python/integrations/fireworks-ai.mdx index 51d00c32a5..64d9fc5bf2 100644 --- a/docs/develop/python/integrations/fireworks-ai.mdx +++ b/docs/develop/python/integrations/fireworks-ai.mdx @@ -252,8 +252,9 @@ the same Worker. The plugin classifies Fireworks AI's HTTP errors. Status 408, 409, 429, and any 5xx retry under the Activity's Retry Policy, and a `Retry-After` header becomes the next retry delay. Connection errors and client-side timeouts retry as well. Other 4xx responses, such as an invalid API key or an unknown model name, fail the Activity without retrying, as -does a model name whose provider has no key on the Worker. Each attempt, its failure, and the final result are in Event -History. +does a model name whose provider has no key on the Worker. Event History records the Activity's result, its attempt +count, and its last failure, not every attempt. Individual failed attempts appear in the Worker's logs and, while the +Activity is still retrying, as the pending Activity's last failure when you describe the Workflow Execution. On serverless, Fireworks AI enforces [adaptive token-per-minute limits](https://docs.fireworks.ai/serverless/rate-limits) per model that grow with steady diff --git a/docs/develop/python/integrations/together-ai.mdx b/docs/develop/python/integrations/together-ai.mdx index 58fb4e8a9e..e9b88b661d 100644 --- a/docs/develop/python/integrations/together-ai.mdx +++ b/docs/develop/python/integrations/together-ai.mdx @@ -72,8 +72,9 @@ each model's ID and whether it supports function calling and structured outputs. `openai/gpt-oss-120b`, is an open-weight reasoning model that supports both; a Workflow names it `together/openai/gpt-oss-120b`. A [dedicated endpoint](https://docs.together.ai/docs/dedicated-endpoints/overview) is addressed the same way: its endpoint string, `/`, goes in the `model` field, and requests -go to `https://api-inference.together.ai/v1` instead of the serverless base URL, so in the sample it is a second entry -in `PROVIDERS`. +go to `https://api-inference.together.ai/v1` instead of the serverless base URL. The sample's `together/` prefix only +reaches the serverless API: to call a dedicated endpoint, add a `PROVIDERS` entry with a distinct name and that base +URL, such as `together-dedicated`, and use its prefix in the model name. A `ModelProvider` routes a `/` name to the matching endpoint and returns the Chat Completions model that the Agents SDK calls: @@ -254,8 +255,9 @@ the same Worker. The plugin classifies Together AI's HTTP errors. Status 408, 409, 429, and any 5xx retry under the Activity's Retry Policy, and a `Retry-After` header becomes the next retry delay. Connection errors and client-side timeouts retry as well. Other 4xx responses, such as an invalid API key or an unknown model name, fail the Activity without retrying, as -does a model name whose provider has no key on the Worker. Each attempt, its failure, and the final result are in Event -History. +does a model name whose provider has no key on the Worker. Event History records the Activity's result, its attempt +count, and its last failure, not every attempt. Individual failed attempts appear in the Worker's logs and, while the +Activity is still retrying, as the pending Activity's last failure when you describe the Workflow Execution. Under high demand, Together AI's serverless API answers `429 Too Many Requests` or `503 Service Unavailable` and asks clients to retry with exponential backoff, which is what the default Activity Retry Policy does; Together AI's