added agent pattern

[gyliu513]

Fixes #2373

Changes

Please provide a brief description of the changes here.

Note: if the PR is touching an area that is not listed in the existing areas, or the area does not have sufficient domain experts coverage, the PR might be tagged as experts needed and move slowly until experts are identified.

Merge requirement checklist

• CONTRIBUTING.md guidelines followed.
• Change log entry added, according to the guidelines in When to add a changelog entry.
  • If your PR does not need a change log, start the PR title with [chore]
• schema-next.yaml updated with changes to existing conventions.

[lmolkova]

I have a bunch of questions:

• who is going to use it and how? There are so many variations of RAG and it can be used within other scenarios, how we're going to classify things that follow multiple patterns
• who is going to set it and how? It does not seem like generic framework instrumentation can set it - there is nothing in the APIs that it maps to. It doesn't seem like individual agents would know if they are used in the multi-agent scenarios.
• why is this scoped to agents? RAG is a pattern used in general LLM scenarios,
• is there a registry of patterns and who maintains it? Are there descriptions of them?

General feedback - we should try to ground attributes to something well understood that has clear use-cases. It could be useful to bring it up for discussion on the SIG call.

[gyliu513]

@PRATIBHA-Moogi can you help comment here? Thanks!

[PRATIBHA-Moogi]

@lmolkova the Agent Style or pattern will help APMs to create custom views to a resources group like VectorDB by means of filtering Agents by used pattern. Currently we can't filter metrics views on VectorDB on Agent basis. To have observability on VectorDB usage and performance when multiple agents using VectorDB as their knowledge source, this attribute is necessary to have which convey which agent is using VectorDB resources.

[PRATIBHA-Moogi]

Some of the known patterns one can apply but also this attribute can use user-entered value for creation of custom dashboards view.

[Copilot]

Pull Request Overview

This PR introduces the new attribute "gen_ai.agent.pattern" to capture the underlying agentic pattern used by GenAI agents. The update spans several files, adding the new attribute to the YAML models, updating the registry documentation, and adjusting reference numbers in the GenAI documentation.

Reviewed Changes

Copilot reviewed 5 out of 5 changed files in this pull request and generated no comments.

Show a summary per file

File	Description
model/gen-ai/spans.yaml	Adds "gen_ai.agent.pattern" with recommended requirement level.
model/gen-ai/registry.yaml	Registers the new attribute with stability, type, and examples.
docs/registry/attributes/gen-ai.md	Introduces the attribute details and updates reference numbering.
docs/gen-ai/gen-ai-agent-spans.md	Adds the attribute in the span documentation and adjusts numbering.
.chloggen/2373.yaml	Changelog entry documenting the added agent pattern attribute.

Comments suppressed due to low confidence (2)

docs/registry/attributes/gen-ai.md:20

• Verify that the reference numbering and associated notes in the documentation remain consistent after adding the new attribute.

| <a id="gen-ai-agent-pattern" href="#gen-ai-agent-pattern">`gen_ai.agent.pattern`</a> | string | The agentic pattern or application style used by the GenAI agent. [1] | `RAG`; `ReAct`; `CoT`; `ToT`; `MultiAgent`; `PlanAndExecute` | ![Development](https://img.shields.io/badge/-development-blue) |

docs/gen-ai/gen-ai-agent-spans.md:87

• Please review the updated reference numbers to ensure they align with the overall documentation conventions after incorporating the new attribute.

**[6] `gen_ai.agent.pattern`:** This attribute captures the underlying agentic pattern or architectural style employed by the GenAI agent. Common patterns include but are not limited to: Retrieval-Augmented Generation (RAG), ReAct (Reasoning and Acting), Chain-of-Thought (CoT), Tree-of-Thoughts (ToT), and other established agentic application patterns. This information is valuable for debugging and performance analysis as different patterns may require different observability approaches and optimizations.