AI-266: Add @temporalio/openai-agents sample suite#479
Conversation
|
Semgrep found 14
Risk: Affected versions of nanoid are vulnerable to Exposure of Sensitive Information to an Unauthorized Actor / Incorrect Type Conversion or Cast. nanoid is vulnerable to information exposure (CVE-2021-23566). Passing a non-integer object with a crafted valueOf() method as the size argument to nanoid(), or to the generators returned by customAlphabet() and customRandom(), tricks the internal random-byte pool into reusing previously generated bytes, allowing a prior ID to be reproduced. Upgrade to nanoid >= 3.1.31. Fix: Upgrade this library to at least version 3.1.31 at samples-typescript/pnpm-lock.yaml:13366. Reference(s): GHSA-qrpm-p2h7-hrv2, CVE-2021-23566 |
Fourteen self-contained samples demonstrating how to run OpenAI Agents SDK agents as Temporal Workflows with the @temporalio/openai-agents integration: the basic building blocks, handoffs, agent patterns, sessions, human approval, hosted/Nexus/MCP tools, tracing, model providers, reasoning content, a research bot, and a customer-service chat. Each sample is a standalone package with a fake-model Worker test that runs without an API key, and the suite is wired into the pnpm workspace, CI, and the samples list.
643db6d to
d33a6dc
Compare
| // Some samples have different config files from those in .shared/ | ||
| // that we don't want to overwrite | ||
| const TSCONFIG_EXCLUDE = [ | ||
| 'openai-agents', |
There was a problem hiding this comment.
Sure you need this? What inside of https://github.com/temporalio/samples-typescript/pull/479/changes#diff-226d2294f138d5c2a050eee0f6ef06ab823a0cb322f2931cf75c9ea480c5f908 needs to be different from the shared one?
| temporal server start-dev | ||
| ``` | ||
|
|
||
| Point the Worker at your OpenAI-compatible endpoint. For example, a local Ollama server: |
There was a problem hiding this comment.
Would it be useful to demonstrate how to start a OpenAI-compatible Ollama server? I assume it would just be a 1-liner
| @@ -0,0 +1,39 @@ | |||
| # OpenAI Agents: Research Bot | |||
|
|
|||
| A multi-agent research Workflow built with `@temporalio/openai-agents`. It mirrors the OpenAI Agents | |||
There was a problem hiding this comment.
The more samples I write, I think most people (and agents) that read samples are typically searching for samples of specific features, not for specific use cases. There are also N features for each plugin and infinite use cases.
What do you think about modifying the title of some of these samples to refer to a feature? (In this case, it looks like we're showing off a multi-agent workflow)
There was a problem hiding this comment.
"customer-service" is another one that could probably use a title change.
Adds fourteen self-contained samples under
openai-agents/— the basic building blocks, handoffs, agent patterns, sessions, human approval, hosted/Nexus/MCP tools, tracing, model providers, reasoning content, a research bot, and a customer-service chat. Each is a standalone package with a fake-model Worker test that runs without an API key, and the suite is wired into the pnpm workspace, CI, and the samples list.