Skip to content

python-api.md

Latest commit

 

History

History
239 lines (210 loc) · 6.18 KB

python-api.md

File metadata and controls

239 lines (210 loc) · 6.18 KB
Raw

Python SDK API

This page covers the stable, high-value Python surface. Import public types from memorizz unless a guide documents a provider-specific module.

MemAgent

::: memorizz.memagent.core.MemAgent options: members: - run - run_stream - validate_configuration - capability_report - semantic_cache_stats - inspect_semantic_cache - invalidate_semantic_cache - generate_summaries - observability_summary - get_trace_context - record_feedback - record_task_outcome - has_automations - create_automation - list_automations - get_automation - pause_automation - resume_automation - trigger_automation - list_automation_runs - delete_automation - list_approval_proposals - approve - reject - cancel_approval - resume_approval - close - lifecycle show_source: false show_root_heading: true

Persistence

agent.save() stores the current definition in its configured memory provider. MemAgent.load(agent_id, memory_provider=provider, **overrides) restores it; omit memory_provider only when the definition lives in the default filesystem provider.

MemAgentBuilder

::: memorizz.memagent.builders.agent_builder.MemAgentBuilder options: members: - with_name - with_instruction - with_model - with_llm_config - with_persona - with_favorite - with_memory_provider - with_memory_ids - with_application_mode - with_tools - with_tool - with_toolbox - with_mcp_servers - with_internet_access_provider - with_skills_marketplace - with_skill_paths - with_sandbox - with_sandbox_provider - with_browser_control - with_browser_control_provider - with_meta_harness - with_execution_harness - with_semantic_cache - with_embedding_provider - with_retrieval_policy - with_context_policy - with_tool_result_policy - with_approval_store - with_skills - with_skill_retrieval - with_skillbox - with_continual_learning - with_learning_control_plane - with_workflow_outcome_evaluator - with_delegation - with_delegates - with_semantic_layer - with_completion_policy - with_automations_enabled - with_default_timezone - with_self_aware - with_entity_memory - with_tool_access - with_max_steps - with_verbose - with_oracle_from_env - with_e2b_from_env - build - build_and_save show_source: false show_root_heading: true

Meta-harness

::: memorizz.metaharness.service.MetaHarness options: members: - from_env - register - list_harnesses - probe - run - start - stream - approve - reject - resume_approval - resume_approval_start - cancel - retry - run_plan - compare - recover_interrupted_runs - close show_source: false show_root_heading: true

::: memorizz.metaharness.models.HarnessTask options: show_source: false show_root_heading: true

::: memorizz.metaharness.base.AgentHarness options: show_source: false show_root_heading: true

See the memory-first meta-harness guide for the adapter security matrix and complete SDK, CLI, UI, and MCP workflows.

Runtime policies

::: memorizz.tooling.ToolPolicy options: show_source: false show_root_heading: true

::: memorizz.tooling.ToolResultPolicy options: show_source: false show_root_heading: true

::: memorizz.tooling.ContextPolicy options: show_source: false show_root_heading: true

::: memorizz.retrieval.RetrievalPolicy options: show_source: false show_root_heading: true

::: memorizz.completion.CompletionPolicy options: show_source: false show_root_heading: true

Model providers

::: memorizz.llms.llm_provider.LLMProvider options: show_source: false show_root_heading: true members: - generate - generate_stream - get_config - get_last_usage - get_context_window_tokens

Memory providers

::: memorizz.memory_provider.base.MemoryProvider options: show_source: false show_root_heading: true members: - store - retrieve_by_query - retrieve_by_id - list_all - retrieve_conversation_history_ordered_by_timestamp - query_observability_records - invalidate_semantic_cache - store_memagent - delete_memagent - list_memagents - close

See the custom provider contract before implementing this interface, particularly the three-state user_id filter.

Memory evaluation

The normalized memory-suite SDK keeps diagnostic runs separate from exact paper reproduction. Use get_protocol_manifest() before a run and inspect the returned report's comparison_label, paper_comparable, and comparability_reasons fields before publishing a comparison.

from memorizz.benchmarks.memory_suite import (
    get_protocol_manifest,
    run_memory_suite,
    verify_dataset,
)

readiness = verify_dataset("longmemeval-v2", data_path="./datasets/lme-v2")
protocol = get_protocol_manifest("longmemeval-v2")

report = run_memory_suite(
    "longmemeval-v2",
    "./datasets/lme-v2",
    variant="standard-100",
    profile="smoke",
    workspace="./.memorizz-eval",
    memory_backend="filesystem",  # use "oracle" for Oracle AI Database
    candidate_pool_size=256,
    lexical_ratio=0.35,
    oracle_reader=True,
)

MemorySuiteRunner additionally accepts injected model, judge, embedding, and memory-provider objects for deterministic tests. See the evaluation-suite guide for official-source sync, strict comparability, CLI equivalents, and result interpretation.