fix(nemo-agents): Support Studio multi-turn transcript replay for Fabric agents

[mmogallapalli]

Summary

• Serialize the complete OpenAI chat transcript into Fabric invocation input for multi-message requests.
• Preserve the existing input exactly for single-message, one-shot requests.
• Retain message roles and ordering so Studio can provide conversational context without managing Fabric session IDs.

Studio already provides multi-turn chat for NAT-backed agents by retaining the conversation client-side and resending the full OpenAI messages transcript on every turn. NAT-backed serving uses that replayed transcript as context rather than relying on a platform-managed session.

This change gives Fabric-backed agents parity with that existing Studio behavior. The Fabric shim translates the replayed transcript into a role-aware input string, while each request remains an independent invocation. Native Fabric session management and session-ID wiring remain separate follow-up work.

Testing

• Added coverage for unchanged single-message input.
• Added coverage for role-aware serialization of system, assistant, and user messages.
• Verified the Fabric server unit suite: 25 passed.
• Verified Ruff and git diff --check pass.

Studio Before:

Studio After:

Summary by CodeRabbit

• Bug Fixes
  • Improved chat interactions by preserving the full conversation history when sending requests.
  • System, assistant, and user messages are now included in the proper order, providing better context for responses.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Enterprise

Run ID: 3c63708d-aa04-4add-9224-803ec3c74791

📥 Commits

Reviewing files that changed from the base of the PR and between 12bfdeb and b4af99a.

📒 Files selected for processing (1)

• plugins/nemo-agents/src/nemo_agents_plugin/fabric/server.py

🚧 Files skipped from review as they are similar to previous changes (1)

• plugins/nemo-agents/src/nemo_agents_plugin/fabric/server.py

---

📝 Walkthrough

Walkthrough

Fabric invocation requests now include the full role-prefixed chat transcript. Unit tests cover single-user messages and multi-turn conversations.

Changes

Fabric transcript serialization

Layer / File(s)	Summary
Serialize complete chat transcripts plugins/nemo-agents/src/nemo_agents_plugin/fabric/server.py, plugins/nemo-agents/tests/unit/test_fabric_server.py	_to_fabric_invocation_request now joins all messages with role prefixes and blank lines. Tests verify single-user and system/assistant/user transcripts.

Possibly related PRs

• NVIDIA-NeMo/nemo-platform#909: Updates _to_fabric_invocation_request and its tests for full transcript serialization.
• NVIDIA-NeMo/nemo-platform#966: Adds the deployed-agent chat bridge that consumes serialized multi-message conversations.

Suggested reviewers: stefan-kickoff

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title clearly and concisely describes the main change: replaying multi-turn Studio transcripts for Fabric agents.
Docstring Coverage	✅ Passed	Docstring coverage is 100.00% which is sufficient. The required threshold is 80.00%.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

✨ Finishing Touches

📝 Generate docstrings

• Create stacked PR
• Commit on current branch

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch mmogallapall/aircore-969-support-studio-transcript-replay-for-fabric-backed-agent

---

_{Comment @coderabbitai help to get the list of available commands.}

[github-actions]

Suite	Lines Covered	Line Rate	Branch Rate
Unit Tests	29440/37448	78.6%	63.2%
Integration Tests	17382/36166	48.1%	20.6%

[tylersbray]

lgtm