agenticmem

/user_profiler/agenticmem/agenticmem_client

Description: Python SDK for remote async access to AgenticMem API

Main Entry Points

• Client: agenticmem/client.py - AgenticMemClient
• Utils: agenticmem/client_utils.py - Helper utilities

Purpose

1. Remote API access - Async SDK for applications to call AgenticMem backend
2. Authentication - Handle login and Bearer token management
3. Type-safe interface - Auto-parsing responses into Pydantic models

API Methods

Authentication:

• login(email, password) - Get auth token

Publishing:

• publish_interaction(request_id, user_id, interactions, source, agent_version) - Publish interactions (triggers profile/feedback/evaluation)

Profiles:

• search_profiles(request) - Semantic search
• get_profiles(request) - Get all for user
• get_all_profiles(limit, status_filter) - Get all profiles across all users
• delete_profile(user_id, profile_id, search_query) - Delete profiles
• get_profile_change_log() - Get history
• rerun_profile_generation(request) - Regenerate profiles from interactions

Interactions:

• search_interactions(request) - Semantic search
• get_interactions(request) - Get all for user
• delete_interaction(user_id, interaction_id) - Delete interaction

Requests:

• get_requests(request) - Get request groups with associated interactions
• delete_request(request_id) - Delete a request and its interactions
• delete_request_group(request_group) - Delete all requests in a group

Feedback:

• get_raw_feedbacks(request) - Raw feedback from interactions
• add_raw_feedback(request) - Add raw feedback directly to storage
• get_feedbacks(request) - Aggregated feedback with status
• rerun_feedback_generation(request) - Regenerate feedback for agent version
• run_feedback_aggregation(request) - Aggregate raw feedbacks into insights

Evaluation:

• get_agent_success_evaluation_results(request) - Get agent success evaluation results

Configuration:

• set_config(config) - Update org config (extractors, evaluators, storage)
• get_config() - Get current config

Architecture Pattern

• All async - Uses aiohttp for HTTP requests
• Type-safe - Pydantic models from agenticmem_commons
• Auto-parsing - Responses → Pydantic models
• Flexible input - Accepts Pydantic models or dicts
• Bearer auth - Automatic token handling