0.2.0

What's Changed

• feat: add anthropic contextual retrieval example by @kemingy in #30
• feat: graph retrieval by @kemingy in #31
• chore: rm breakpoints in the code by @kemingy in #32
• feat: use gemini API to extract the PDF directly by @kemingy in #33
• feat: change to async interface by @kemingy in #35
• fix: encoder hook for numpy array by @kemingy in #36
• docs: update examples and docs to use the async interface by @kemingy in #37
• feat: support run defined pipeline from config by @kemingy in #38
• fix: dynamic pipeline namespace & keyword args by @kemingy in #39
• feat: add rate limit to gemini client by @kemingy in #40
• feat: support voyage multimodal embedding by @kemingy in #41
• feat: use psycopg pool to avoid conflicts in concurrent request by @kemingy in #42
• fix: split text-emb & multimodal-emb by @kemingy in #43
• refact: use msgspec struct to simplify gemini req/resp construction by @kemingy in #44
• chore: add codeql.yml by @kemingy in #45
• feat: add dockerfile, push to ghcr on release by @kemingy in #46
• refact: add provider as base class to reduce duplicate code by @kemingy in #47
• chore: fix the pages build step permission by @kemingy in #48
• fix: resolve some type errors by @kemingy in #49
• feat: add jina emb by @kemingy in #50
• refact: reduce the complexity of voyage client by @kemingy in #51
• feat: add UMBRELA LLM eval metric by @kemingy in #53
• feat: support graph index in dynamic pipeline by @kemingy in #52
• feat: Add llamaparse external ocr service support by @xieydd in #54
• docs: add dynamic pipeline example by @kemingy in #55
• feat: jina multimodal emb by @kemingy in #56
• feat: support eval in dynamic pipeline by @kemingy in #57
• fix: pad the defaults to match the length of fields by @kemingy in #58
• fix: add a fake chunk for graph on image/pdf by @kemingy in #59
• feat: allow user prompt by @kemingy in #60
• feat: support eval with image by @kemingy in #61

New Contributors

• @xieydd made their first contribution in #54

Full Changelog: 0.1.0...0.2.0