- Add embedding search methods (#293)
- Batch embedding pipeline (#292)
- Add task_id filtering to get_unprocessed_threads (#290)
- Store error count on ETLTask (#285)
- Add per-item model validation in pipes (#283)
- Only store llm generated asset description (#287)
- Reuse validated_items helper in google pipe (#286)
- Parse IG comments with gif as comment (#284)
- Get supported MIME from base llm client (#288)
- Asset descriptions pipeline (#280)
- Add thread.content for semantic representation (#277)
- Add autogen pipe for telegram conversations (#271)
- Use image.name as create content (#279)
- Only consider memories batches for memories (#278)
- Vertex ai lite llm gcs config (#275)
- Basic memory pipeline for google (#273)
- Do not bound memory generation (#272)
- Add snapshot tests to remaining IG Pipes (#266)
- [breaking] Store threads before callback (#269)
- [breaking] Return inserted thread ids from
insert_threads(#268) - (proxy) Allow clients to control enrichment (#267)
- Add recent threads from collection to group context (#264)
- [breaking] Add opt collection_id to thread (#262)
- [breaking] Require credentials and bucket name in
VertexAIConfig(#263)
- (proxy) [breaking] Switch
hostheader to<prefix>-upstream-host(#260) - (proxy) Allow to specify the ctxuse headers prefix (#259)
- [breaking] Only support
gemini-2.5-profor vertex ai models (#252) - [breaking] Require embedding dimension in to init store (#251)
- [breaking] Switch to litellm base client (#247)
- Cleanup (#257)
- Extract GroupContextBuilder (#255)
- Change BasePromptBuilder to single GroupContext interface (#254)
- (proxy) Ignore encoding from client (#249)
- Import configs utils (#248)
- Move local ctx setup in cli (#245)
- Added new available providers and commands in readme (#246)
- Autogen pipe for instagram_ads_clicked interaction type (#243)
- Autogen schemas for v0/v1 story_likes on IG (#240)
- [breaking] Memory service used by both agent and cli (#241)
- Add instagram_ads_viewed pipe and FibreAd models (#235)
- Split pipes for story_likes and liked_posts (#239)
- Payload test for videos_watched (#224)
- Do not package json schemas (#242)
- [breaking] Post response proxy callback (#236)
- [breaking] Move agent inside
ContextUse(#234)
- Add /v1/responses API proxy support alongside /v1/chat/completions (#228)
- Proxy rich logging (#231)
- Skip release commits in changelog (#232)
- Generate session id if not set in the header (#229)
- Add
--upstream-urlto the proxy command (#225) - Filters for memory generation (#226)
- Context proxy as first class citizen (#221)
- Semantic facets for memories (#222)
- (test) Add payload test to pipe test kit (#223)
- Debug log request and more details on memories ops (#227)
- Add Netflix provider with 6 interaction types (#209)
- Add Airbnb provider including additional scopes (#210)
- Resolve asset_uri in pipe.run (#217)
- Use InstagramBaseModel in media pipe (#219)
- (chatgpt) Update schema based on different types of archives (#215)
- Group dependencies changes in changelog (#214)
- Proxy stores threads and generates memories in the background (#206)
- Add user profile as context to memory generation prompt (#205)
- Upsert memory embeddings (#207)
- [breaking] Downgrade
google-adkto >=1.22 (#212)
- Move tests using llm to evals (#208)
- Basic memory unit and integration tests (#204)
- Banner on root and help commands (#199)
- Context proxy (#193)
- Claude conversations with autogen schema guidelines (#194)
- Memories from agent and human conversations (#177)
- (proxy) Do not enrich request if
max_tokensis below a threshold (#202) - Parse ig story likes with v0 pipe (#185)
- Update ig etls according to guidelines (#180)
- Chatgpt conversations with autogen schema (#198)
- Google interactions with autogen schema (#195)
- Split pipes, schemas and records into separate files (#189)
- Chatgpt etl according to guidelines (#188)
- Input file items schema validation for google (#181)
- Input file items schema validation for claude (#186)
- Input file items schema validation for chatgpt (#183)
- Input file schema validation for ig (#182)
- (test) Uniform provider etls testing (#175)
- Add demo to readme (#201)
- Guidelines for automatic schema generation from exports (#192)
- Update guidelines for input schema validation (#187)
- Simplify provider guidelines for new etls (#178)
- Skill.md ai optimization (#176)
- Stream ig interactions in etl (#184)
- Update readme (#191)
- Etl to process IG DMs (#173)
- Memory batch status spinners (#170)
- Agents skill (#165)
- Improve cli startup speed (#167)
- Check build is working across python versions (#169)
- Simplify agents.md and extract the guide to add a provider (#164)
- Remove data folder placeholder (#163)
- Bring back future annotations (#166)
- [breaking] Only require zip path on quick mode (#151)
- Add ETL for instagram_posts interaction type (#159)
- [breaking] Change data folder to
context-use-data(#152) - Memories commands prompt for api key if not set (#147)
- Add --version command to cli (#144)
- Generalize archive path pattern for post comments (#160)
- Rename data folder where missing (#158)
- Do not sleep after batch on quickstart (#150)
- Remove future annotations imports where not needed (#154)
- Update readme (#161)
- Run check PR title on all PR events (#157)
- Pypi page details (#155)
- Ensure conventional commits (#156)
- Update uv.lock on prepare-release (#145)
- [breaking] Remove mcp (#138)
- [breaking] Adk is now required (#136)
- [breaking] Remove quickstart command (#134)
- Make db path configurable (#133)
- [breaking] Remove in-memory store (#132)
- [breaking] Remove postgres (#131)
- [breaking] Remove set-store config command (#130)
- [breaking] Use sqlite for all commands (#129)
- [breaking] Sqlite store as default (#128)
- Update readme (#141)
- Remove
vprefix from generate version (#140) - Mit license (#142)
- Update readme (#137)
- Make sure we don't bump major when major is 0 (#135)
First release of context-use.