Pull requests · deepset-ai/haystack

fix: reject non-mapping input when deserializing a Pipeline

#13739zhengqiuyang opened

Nov 13, 2026

1

fix: reject incomplete summaries in SummarizationCompactor

#13737HYYAlex opened

Nov 13, 2026

2

fix: preserve non-string dict keys in pipeline snapshot serialization

#13732Bensonluo opened

Nov 13, 2026

7 tasks done

2

fix: convert common Python types in Agent tool results to JSON

#13728astroknot-sheep opened

Nov 13, 2026

Contributor

8

fix: preserve Unicode in recursive token splitting

#13721wkd09 opened

Nov 12, 2026

3 of 6 tasks

3

perf: avoid quadratic start check when streaming from OpenAIResponsesChatGenerator

#13716deadstrobe5 opened

Nov 12, 2026

3

fix: keep multi-byte characters intact when RecursiveDocumentSplitter splits by token

#13710sunhouy opened

Nov 12, 2026

Contributor

2

fix: close remaining pipeline components when one close() fails

#13707kevin9327 opened

Nov 12, 2026

13

fix: propagate caller cancellation from PipelineStreamHandle.aclose()

#13703inchang-ing opened

Nov 11, 2026

Contributor5 tasks done

2

fix: end `Pipeline.stream` iteration when `aclose` cancels the pipeline before it starts

#13702prassu017 opened

Nov 11, 2026

4

fix: validate ground-truth relevance scores in DocumentNDCGEvaluator

#13697jjhong011001 opened

Nov 11, 2026

3 tasks done

2

fix: avoid whitespace-only and over-long chunks in RecursiveDocumentSplitter

#13693MaiAditya opened

Nov 11, 2026

3

fix(tools): keep dataclass default factories lazy

#13690Elon-Jiang opened

Nov 11, 2026

7 tasks done

4

fix: AnswerBuilder returns whole match when optional capture group does not participate

#13677zuizhuai opened

Nov 11, 2026

Contributor

2

feat: add remove_extra_whitespaces option to TextCleaner

#13672HareeshDaxton opened

Nov 10, 2026

Contributor7 tasks done

4

fix: serialize pydantic models in JSON mode so snapshots with datetime fields can be saved

#13665zzx04-cool opened

Nov 10, 2026

7 tasks done

8

fix: handle LLMEvaluator chat generator replies without text

#13664AAliKKhan opened

Nov 10, 2026

5

fix: let Pipeline.connect accept a Literal output for an input of its value type

#13642aryamangodara opened

Nov 9, 2026

Contributor

3

feat: Add SemanticResolver component for entity resolution using TEI (#10985)

#13627tech00exploere opened

Nov 9, 2026

7 tasks done

2

fix: name a tool output handler that has no name

#13622lxcxjxhx opened

Nov 8, 2026

3

fix: raise FilterError for incomparable document values in range filters

#13618Harshil26202 opened

Nov 7, 2026

2

docs: correct MarkdownHeaderSplitter content exception types

#13592Ymmcc opened

Nov 6, 2026

5

docs: make the docs site readable by AI agents

#13588agnieszka-m opened

Nov 6, 2026

Contributor

3

fix: preserve falsy non-mapping Markdown frontmatter

#13569flysheep-ai opened

Nov 5, 2026

4 of 6 tasks

4

docs: remove dead Haystack 1.x docs ZIP links and fix template example link

#13568leom1919 opened

Nov 5, 2026

5

Navigation Menu

Repository navigation

All pull requests

fix: reject non-mapping input when deserializing a Pipeline

fix: reject incomplete summaries in SummarizationCompactor

fix: preserve non-string dict keys in pipeline snapshot serialization

fix: convert common Python types in Agent tool results to JSON

fix: preserve Unicode in recursive token splitting

perf: avoid quadratic start check when streaming from OpenAIResponsesChatGenerator

fix: keep multi-byte characters intact when RecursiveDocumentSplitter splits by token

fix: close remaining pipeline components when one close() fails

fix: propagate caller cancellation from PipelineStreamHandle.aclose()

fix: end `Pipeline.stream` iteration when `aclose` cancels the pipeline before it starts

fix: validate ground-truth relevance scores in DocumentNDCGEvaluator

fix: avoid whitespace-only and over-long chunks in RecursiveDocumentSplitter

fix(tools): keep dataclass default factories lazy

fix: AnswerBuilder returns whole match when optional capture group does not participate

feat: add remove_extra_whitespaces option to TextCleaner

fix: serialize pydantic models in JSON mode so snapshots with datetime fields can be saved

fix: handle LLMEvaluator chat generator replies without text

fix: let Pipeline.connect accept a Literal output for an input of its value type

feat: Add SemanticResolver component for entity resolution using TEI (#10985)

fix: name a tool output handler that has no name

fix: raise FilterError for incomparable document values in range filters

docs: correct MarkdownHeaderSplitter content exception types

docs: make the docs site readable by AI agents

fix: preserve falsy non-mapping Markdown frontmatter

docs: remove dead Haystack 1.x docs ZIP links and fix template example link