All notable changes to the CueMap Python SDK will be documented in this file.
- Multi-Hop Recall:
depthparameter in recall requests to enable multi-hop associative retrieval.
- Optional alias expansion: Added optional alias expansion to the SDK.
- Context Expansion: New
context_expandmethod to retrieve related concepts from the cue graph. - Cloud Backup Management: New methods (
backup_upload,backup_download,backup_list,backup_delete) to manage cloud snapshots programmatically.
- Ingestion API: New methods
ingest_url,ingest_content, andingest_filefor direct content ingestion. - Lexicon Management: New methods (
lexicon_wire,lexicon_inspect,lexicon_graph,lexicon_synonyms,lexicon_delete) for manual control over the engine's associative graph. - Job Status: New
jobs_status()method to track background ingestion progress. - Brain Control Flags: Added parameters to disable specific brain modules (
disable_pattern_completion,disable_salience_bias, etc.) for deterministic debugging.
- BREAKING: Refactored
recallmethod signature.query_textis now the first positional argument, followed bycuesandprojects, to prioritize Natural Language Search.- Old:
client.recall(cues=["tag"], query_text="search") - New:
client.recall("search", cues=["tag"])
- Old:
- Documentation: Updated README to reflect the "Brain-Inspired" architecture and new API surface.
- Asynchronous Client: Introduced
CueMapAsyncClientfor high-performance non-blocking ingestion and recall. - Explainable Recall: Added
explain: boolparameter torecall_weightedandrecallmethods to retrieve detailed scoring metadata. - Advanced Search Parameters: Support for
min_intersectionandauto_reinforcein recall requests. - Alias Management: New methods to add, list, and delete cue aliases for semantic boosting.
- Grounding Utilities: Built-in methods to filter and select the most relevant memories from recall results.
- Multi-tenant Support: Updated
ProjectConfigto support project-specific normalization and taxonomy settings.
- Default Weights: Unified the weighting system to match the Rust engine's continuous gradient scoring.
- Connection Handling: Improved retry logic and timeout management for robust communication with the engine.
- UTF-8 Handling: Ensured consistent encoding when ingesting non-English or special-character content.
- JSON Parsing: Added robust Regex fallback for parsing LLM-generated JSON responses.
- Synchronous
CueMapClientstable release. - Integration with basic multi-project engine features.
- Performance monitoring hooks.
- Standardized REST-based communication.
- Basic CLI for SDK interaction.
- Unit test suite for core client logic.
- Persistence stubs and local caching.
- Enhanced error reporting and retry strategies.
- Initial Python binding prototype.
- Basic memory ingestion and search methods.
Note: This version is designed to work with CueMap Rust Engine v0.6.x.