Skip to main content

Session Management

Function Introduction​

Dakou is an asynchronous Agent product. Users can delegate tasks to Dakou for processing at any time via mobile/PC. After delegating tasks to Dakou, users do not need to wait online. Dakou processes tasks in the background sandbox environment independently, and users only need to review the results after Dakou completes execution.

Detailed Explanation​

Delegate Tasks​

View Recent Sessions​

View All Historical Sessions​

  • Operations on sessions include:
    • Favorite: For sessions that need focused attention or future review, they can be favorited for quick retrieval
    • Rename: Currently, session names are summarized by AI, but users can customize session names

Usage Recommendations​

Task Decomposition​

Break down complex tasks into multiple relatively independent smaller tasks

  • For Agents:
    • Smaller tasks have clearer intentions, making Agents more focused and resulting in better output
    • LLM large models have context window limitations, and task decomposition avoids exceeding these limits
  • For Users: Starting with smaller tasks helps explore Agent capability boundaries, enabling better Agent usage
  • Example: Break down complex requirements into modular tasks

Clear Description and Success Criteria​

Clearly define task descriptions and completion standards

  • For Agents:
    • Clear task descriptions help Agents understand task scope and requirements, avoiding ineffective expansion during processing
    • Clear completion standards help Agents know when tasks are finished, avoiding dead loops or deviations from expectations
  • For Users: Clear task descriptions and completion standards enable Agents to process more accurately and efficiently, avoiding resource waste
  • Example:
    • Task Description: Develop an AI Agent website using a black theme
    • Completion Standard: Need to create a code review

Define Clear Validation Standards​

  • For Agents: Clear validation standards help Agents verify whether task results meet requirements, facilitating iterative improvements
  • For Users: Defining validation standards makes Agent processing results more in line with expectations, avoiding ineffective multi-round interactions
  • Example: In generating unit test cases, unit test coverage should reach 80%

Parallel Processing​

Parallel processing, asynchronous execution, no need to wait

  • For Users:
    • Delegating multiple tasks to Agents for parallel processing results in higher overall task processing efficiency
    • Agents will launch containers to asynchronously execute user-delegated tasks, eliminating the need for users to wait online

Make Good Use of Playbooks​

Playbooks are reusable and shareable Prompts that facilitate the replication of successful experiences

  • For Agents: Can obtain clearer task requirements, resulting in results more in line with user expectations
  • For Users:
    • Playbooks are reusable, simplifying user input for similar problems
    • Playbooks are shareable, facilitating the replication of successful experiences for specific problems
  • Example: Standard SOPs for specific problem types