Skip to content

Conversation

@JeffreySu
Copy link
Contributor

…to improve the accuracy of name hits

Improvement content:

  1. Enforce the use of the JSON standard return format.
  2. Introduce a one-time retry mechanism to reduce the error rate (currently, all tests pass 100%).

Checks

…to improve the accuracy of name hits

Improvement content:
1. Enforce the use of the JSON standard return format.
2. Introduce a one-time retry mechanism to reduce the error rate (currently, all tests pass 100%).
…to improve the accuracy of name hits

Improvement content:
1. Enforce the use of the JSON standard return format.
2. Introduce a one-time retry mechanism to reduce the error rate (currently, all tests pass 100%).
@LittleLittleCloud LittleLittleCloud self-requested a review June 17, 2025 16:39
var name = response.GetContent() ?? throw new ArgumentException("No name is returned.");
var responseMessageStr = response.GetContent() ?? throw new ArgumentException("No name is returned.");

RolePlayOrchestratorResponse? responseMessage;
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

nit:

var responseMessage = JsonSerializer.Deserialize<RolePlayOrchestratorResponse>(responseMessageStr) ?? throw new InvalidOperationException("Incorrect RolePlayOrchestratorResponse JSON format.");

var reaginCandidate = candidates.FirstOrDefault(x => x.Name!.ToUpper() == regainResponseMessage.Speaker!.ToUpper());

if (reaginCandidate != null)
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

reaginCandidate -> regainCadidate

namespace AutoGen.Core.Orchestrator;
internal class RolePlayOrchestratorResponse
{
internal string? Speaker { get; set; }
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The access modifier of property doesn't have to be also internal here?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

public is also fine.

Each message will start with 'From name:', e.g:
From {agentNames.First()}:
//your message//.");
## Available Speaker Names
Copy link
Collaborator

@LittleLittleCloud LittleLittleCloud Jun 17, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Maybe adding a class-level summary on top?

/// <summary>
/// This orchestrator uses a robust two-step strategy to select the next speaker in a roleplay conversation:
/// 1. It first prompts the LLM to select the next speaker from the list of valid candidate names, requiring output in a strict JSON format.
/// 2. If the LLM's chosen name does not exactly match any candidate (e.g., due to hallucination, abbreviation, or formatting issues), 
///    the orchestrator issues a second prompt to the LLM, instructing it to map the provided name to the closest valid candidate name from the original list.
/// This approach ensures that the selected speaker always corresponds to an authorized candidate and guards against LLM output errors.
/// </summary>
Copy link
Collaborator

@LittleLittleCloud LittleLittleCloud left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

PR LGTM, minor changes requested before merging

@codecov
Copy link

codecov bot commented Jun 17, 2025

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 79.71%. Comparing base (89927ca) to head (8912db2).

Additional details and impacted files
@@           Coverage Diff           @@
##             main    #6688   +/-   ##
=======================================
  Coverage   79.71%   79.71%           
=======================================
  Files         232      232           
  Lines       17323    17323           
=======================================
  Hits        13809    13809           
  Misses       3514     3514           
Flag Coverage Δ
unittests 79.71% <ø> (ø)

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:
  • ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
  • 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

2 participants