Add AND-capable trigger conditions to Python SummarizationMiddleware (parity with LangChain.js)

[yasirali0]

Checked other resources

• This is a feature request, not a bug report or usage question.
• I added a clear and descriptive title that summarizes the feature request.
• I used the GitHub search to find a similar feature request and didn't find it.
• I checked the LangChain documentation and API reference to see if this feature already exists.
• This is not related to the langchain-community package.

Package (Required)

• langchain
• langchain-openai
• langchain-anthropic
• langchain-classic
• langchain-core
• langchain-cli
• langchain-model-profiles
• langchain-tests
• langchain-text-splitters
• langchain-chroma
• langchain-deepseek
• langchain-exa
• langchain-fireworks
• langchain-groq
• langchain-huggingface
• langchain-mistralai
• langchain-nomic
• langchain-ollama
• langchain-perplexity
• langchain-prompty
• langchain-qdrant
• langchain-xai
• Other / not sure / general

Feature Description

In LangChain Python, SummarizationMiddleware(trigger=...) currently supports: (https://reference.langchain.com/python/langchain/middleware/#langchain.agents.middleware.SummarizationMiddleware)

• a single ContextSize tuple like ("tokens", 3000) or ("messages", 50)
• or a list of tuples, where summarization runs when any threshold is met (OR).

In LangChain.js, trigger supports AND conditions within a single trigger object (e.g. tokens >= 4000 AND messages >= 10), and OR across multiple trigger objects. https://reference.langchain.com/javascript/functions/langchain.index.summarizationMiddleware.html

I’d like Python SummarizationMiddleware(trigger=...) to support the same expressiveness (AND + OR), while remaining fully backward compatible.

Use Case

I want to avoid summarizing too early when only one metric spikes transiently.
Example: summarize only when both tokens and messages are high, e.g.

• tokens >= 4000 AND messages >= 10
  Also want combined OR cases, e.g.
• (tokens >= 5000 AND messages >= 3) OR (tokens >= 3000 AND messages >= 6)

Proposed Solution

Extend Python’s trigger parameter to accept either the existing tuple form or a dict-style “conjunction” form:

1. Keep existing behavior unchanged:
   • trigger=("messages", 50)
   • trigger=[("fraction", 0.8), ("messages", 100)] # OR semantics remains as documented
2. Add AND-capable trigger object(s):
   • trigger={"tokens": 4000, "messages": 10} # AND across keys
   • trigger=[{"tokens": 5000, "messages": 3}, {"tokens": 3000, "messages": 6}] # OR across dicts
3. Implementation sketch (conceptual)
   Normalize trigger into a list of “clauses”, where each clause is a dict of thresholds:
   • tuple ("tokens", 3000) ⇒ {"tokens": 3000}
   • dict {"tokens": 4000, "messages": 10} ⇒ as-is
     Then evaluate:
   • a clause matches when all thresholds in that clause are met (AND)
   • summarization triggers when any clause matches (OR)
     This exactly matches LangChain.js semantics.

Alternatives Considered

No response

Additional Context

No response

[Pobitro-B]

Hi! I would love to take this up. The proposed solution looks good and I am planning to implement it and keep backwards compatibility. Happy to modify based on feedback.