0.0.4

[0.0.4] - 2026-02-25

Added

• on_before_compress callback on ContextManagerMiddleware — called with
  (messages_to_discard, cutoff_index) before compression summarizes and discards
  messages. Enables persistent history archival (e.g. save full conversation to
  files before pruning).
• on_after_compress callback — called with compressed messages after
  compression. Return a string to re-inject it into context as a SystemPromptPart
  (inspired by Claude Code's SessionStart hook with compact matcher).
• Continuous message persistence via messages_path on ContextManagerMiddleware —
  every message (user input, agent responses, tool calls) is saved to a single
  messages.json file on every history processor call. On compression, the summary
  is appended to the same file. The file is the permanent, uncompressed record of
  the full conversation. Supports session resume (loads existing history on init).
• Guided compaction — _compress() and _create_summary() accept a focus
  parameter (e.g., "Focus on the API changes") appended to the summary prompt.
• request_compact(focus) method — request manual compaction on the next
  __call__, with optional focus instructions.
• compact(messages, focus) method — directly compact messages with LLM
  summarization (for CLI /compact commands).
• max_tokens auto-detection from genai-prices — when max_tokens=None
  (the new default), the middleware resolves the model's context window
  automatically via genai-prices. Falls back to 200,000 if not found.
• resolve_max_tokens(model_name) function exported from the package —
  standalone lookup of context windows from genai-prices.
• model_name parameter on ContextManagerMiddleware and factory — used for
  auto-detection of max_tokens when not explicitly set.
• Async token counting — TokenCounter type now accepts both sync and async
  callables (Callable[..., int] | Callable[..., Awaitable[int]]). Enables use of
  provider token counting APIs (e.g. Anthropic's /count_tokens endpoint) or
  pydantic-ai's count_tokens() method. (#6)
• async_count_tokens() helper function exported from the package.
• BeforeCompressCallback, AfterCompressCallback type aliases exported.
• messages_path, model_name, on_before_compress, on_after_compress
  parameters added to create_context_manager_middleware() factory.
• Examples — 6 runnable examples in examples/ covering all features:
  auto-compression, persistence, callbacks, auto-detection, interactive chat,
  standalone processors.

Changed

• max_tokens default changed from 200_000 to None (auto-detect from
  genai-prices, fallback to 200,000).
• keep default changed from ("messages", 20) to ("messages", 0) —
  on compression, only the LLM summary survives (like Claude Code). This produces
  the most compact context after compression.
• Validation now allows 0 for messages/tokens keep and trigger values
  (previously required > 0). Negative values are still rejected.

Dependencies

• genai-prices used for auto-detection of context windows (already a transitive
  dependency via pydantic-ai-middleware).

0.0.3

[0.0.3] - 2025-02-15

Added

• ContextManagerMiddleware - Dual-protocol middleware for real-time context management

  • Acts as pydantic-ai history_processor for token tracking and auto-compression
  • Acts as pydantic-ai-middleware AgentMiddleware for tool output truncation
  • on_usage_update callback for real-time usage tracking
  • max_tool_output_tokens for limiting individual tool output sizes
  • create_context_manager_middleware() factory function
  • Requires hybrid extra: pip install summarization-pydantic-ai[hybrid]

• Shared cutoff algorithms - New internal _cutoff.py module extracted from processors

  • validate_context_size() - Context configuration validation
  • should_trigger() - Trigger condition evaluation
  • determine_cutoff_index() - Retention-aware cutoff calculation
  • find_safe_cutoff() - Tool call/response pair preservation
  • find_token_based_cutoff() - Binary search token cutoff
  • is_safe_cutoff_point() - Safety validation for cutoff points
  • validate_triggers_and_keep() - Configuration normalization
  • Reduces code duplication between SummarizationProcessor and SlidingWindowProcessor

• ModelType type alias (str | Model | KnownModelName) exported from the package for convenience.

Changed

• Lightweight dependency: Replaced pydantic-ai with pydantic-ai-slim to avoid pulling in unnecessary model-specific SDKs (openai, anthropic, etc.). (#4)
• Custom model support: SummarizationProcessor.model, ContextManagerMiddleware.summarization_model, and factory functions now accept str | Model | KnownModelName — enabling custom providers like Azure OpenAI. (#3)
• Code refactoring: Extracted common logic from processor.py and sliding_window.py into shared _cutoff.py module
• README: Updated with ContextManagerMiddleware, hybrid extra, and new features

Dependencies

• Added hybrid extra: pydantic-ai-middleware>=0.2.0 (optional)
• pydantic-ai-middleware added to dev dependencies
• Replaced pydantic-ai>=0.1.0 with pydantic-ai-slim>=0.1.0

0.0.2

[0.0.2] - 2025-01-22

Changed

• README: Complete rewrite with centered header, badges, Use Cases table, and vstorm-co branding
• Documentation: Updated styling to match pydantic-deep pink theme
  • Inter font for text, JetBrains Mono for code
  • Pink accent color scheme
  • Custom CSS and announcement bar
• mkdocs.yml: Updated with full Material theme configuration

Added

• Custom Styling: docs/overrides/main.html, docs/stylesheets/extra.css
• Abbreviations: docs/includes/abbreviations.md for markdown expansions

0.0.1-fix

[0.0.1] - 2025-01-20

Added

• SummarizationProcessor - History processor that uses LLM to intelligently summarize older messages when context limits are reached

  • Configurable triggers: message count, token count, or fraction of context window
  • Configurable retention: keep last N messages, tokens, or fraction
  • Custom token counter support
  • Custom summary prompt support
  • Safe cutoff detection - never splits tool call/response pairs

• SlidingWindowProcessor - Zero-cost history processor that simply discards old messages

  • Same trigger and retention options as SummarizationProcessor
  • No LLM calls - instant, deterministic processing
  • Ideal for high-throughput scenarios

• Factory functions for convenient processor creation:

  • create_summarization_processor() - with sensible defaults
  • create_sliding_window_processor() - with sensible defaults

• Utility functions:

  • count_tokens_approximately() - heuristic token counter (~4 chars per token)
  • format_messages_for_summary() - formats messages for LLM summarization

• Type definitions:

  • ContextSize - union type for trigger/keep configuration
  • ContextFraction, ContextTokens, ContextMessages - specific context size types
  • TokenCounter - callable type for custom token counters

• Documentation:

  • Full MkDocs documentation with Material theme
  • Concepts, examples, and API reference
  • Integration examples with pydantic-ai