Commit Graph

1720 Commits

Author SHA1 Message Date
Matthew Zhou
5337e5bcac feat: Move message embedding to background task [LET-4189] (#4430)
* Test background message embedding

* Change to content
2025-09-04 15:05:35 -07:00
Shubham Naik
d23318b802 chore: fix backend error fix (#4428)
Co-authored-by: Shubham Naik <shub@memgpt.ai>
2025-09-04 14:10:32 -07:00
Kian Jones
2eb7da9bc3 chore: Move core tests into apps/core (OSS Migration) [LET-4169] (#4376)
* add a bunch of test to oss

* symlink and auto-detect dir

* symlink the other direction

* add pull_request_target logic

* remove undertaker and add alembic validation

* symlink doesn't work with gh actions and add validation workflow to ensure actions in cloud and oss are lockstep

* sync these

* specify extras selectively
2025-09-04 13:46:11 -07:00
Matthew Zhou
def95050e2 feat: Add ranks to archival memory search [LET-4193] (#4426)
* Add ranks to archival memory search

* Fix test managers

* Fix archival memory test
2025-09-04 13:35:54 -07:00
Matthew Zhou
ef225d3e49 fix: Temp remove timezone test (#4427)
Temp remove test
2025-09-04 12:05:16 -07:00
Shubham Naik
f6ec6dc3a0 feat: add the ability to find/delete by deployment id [PRO-1150] (#4421)
* feat: add the ability to find/delete by deployment id

* fix: add ability to delete by deployment id

---------

Co-authored-by: Shubham Naik <shub@memgpt.ai>
2025-09-04 11:08:54 -07:00
Shubham Naik
eb21668f35 feat: add hidden property to group and blocks [PRO-1145] (#4415)
* feat: add hidden property to group and blocks

* feat: add hidden property to group and blocks

* chore: bup

* chore: add hidden property

* chore: next

---------

Co-authored-by: Shubham Naik <shub@memgpt.ai>
2025-09-04 10:53:16 -07:00
Shubham Naik
a65400af84 chore: surface the templates rest endpoint (#4412)
* chore: surface the templates rest endpoint

* chore: fix python server stuff

---------

Co-authored-by: Shubham Naik <shub@memgpt.ai>
2025-09-04 08:48:46 -07:00
Sarah Wooders
9e20a76d2a feat: use datetime ranges for archival memory query (#4398) 2025-09-03 20:08:10 -07:00
Matthew Zhou
d924cc005b fix: change to pure rank-based RRF for relevance ordering (#4411)
* Fix RRF

* Fix turbopuffer tests
2025-09-03 17:33:19 -07:00
Kian Jones
fc50a41680 chore: add back dependencies for desktop (#4409)
add back dependencies for desktop
2025-09-03 17:12:04 -07:00
Matthew Zhou
2af6049d6f feat: Change web search to exa [LET-4190] (#4401)
* Change web search to exa

* Fix tf/justfile

* add exa api key for integration test

* Mock exa

---------

Co-authored-by: Kian Jones <kian@letta.com>
2025-09-03 15:52:10 -07:00
jnjpng
12d8242d94 fix: retry on 500 and 503 for gemini [LET-4185]
* handle 500 and 503

* timeout

---------

Co-authored-by: Letta Bot <noreply@letta.com>
2025-09-03 14:02:58 -07:00
Charles Packer
dd1a9a0b7c fix: reduce the default tool return to something more sane [LET-4180] (#4383)
fix: reduce the default tool return to something more sane (prev limit was allowing like 200k tokens to come through and crash gpt-5
2025-09-03 13:51:33 -07:00
Cameron Pfiffer
6438bc5168 fix: correct messages API method from send to create (#4299)
Updated documentation examples to use the correct API method:
- Changed groups.messages.send() to groups.messages.create()
- Changed agents.messages.send() to agents.messages.create()
- Updated parameter format to use messages array with role/content

Fixes inconsistent API usage across documentation and examples.
2025-09-03 13:44:12 -07:00
Matthew Zhou
129dd97902 feat: Add fetch webpage tool [LET-4188] (#4395)
* Add fetch webpage tool

* Use trafilatura for web extraction
2025-09-03 13:34:35 -07:00
Matthew Zhou
051a5cde6a feat: Add archival search endpoint [LET-4184] (#4390)
* Add archival search endpoint

* Run fern autogen

* Add de-dupe logic
2025-09-03 10:55:20 -07:00
Matthew Zhou
6e633bd8f9 feat: Change namespace to be org scoped and filter on agent_id [LET-4163] (#4368)
* Change to org scoped and agent_id filtering

* Finish modifying conversation search tool

* Fix failing tests

* Get rid of bad imports
2025-09-03 10:00:19 -07:00
Charles Packer
507784f292 fix: try and patch the PATCH/update issue with MCP server URL [LET-3933]
* fix: patch update path

* autogenerate

* update

* update

---------

Co-authored-by: Jin Peng <jinjpeng@Jins-MacBook-Pro.local>
Co-authored-by: Letta Bot <noreply@letta.com>
Co-authored-by: jnjpng <jin@letta.com>
2025-09-03 09:42:57 -07:00
cthomas
93ef177ab6 feat: always default to openai client in llm client factory [LET-4176] (#4377)
feat: always default to openai client in llm client factory
2025-09-03 09:40:59 -07:00
jnjpng
cfe47489d8 fix: clean property schema reference
fix

Co-authored-by: Letta Bot <noreply@letta.com>
2025-09-03 08:42:46 -07:00
Charles Packer
1ae0ab1790 fix: properly throw context window exceeded error on the new style of… [LET-4179] (#4380)
fix: properly throw context window exceeded error on the new style of context window overflow from gpt-5
2025-09-02 22:20:46 -07:00
Charles Packer
361c50bce4 fix: patch context window of gpt-5 to 272k not 400k [LET-4177] (#4379)
fix: patch context window of gpt-5 to 272k not 400k
2025-09-02 21:10:40 -07:00
jnjpng
a9e20097bd fix: mcp tool schema formatting for anthropic streaming [LET-4165]
* fix

* update commnet

---------

Co-authored-by: Letta Bot <noreply@letta.com>
2025-09-02 20:44:43 -07:00
cthomas
11d2507877 feat: refactor streaming route logic (#4369) 2025-09-02 19:27:13 -07:00
Charles Packer
bcfd88ae98 fix: patch cache bust in the system prompt via timestamp [LET-4155] (#4375)
fix: patch cache bust in the system prompt via timestamp
2025-09-02 18:07:13 -07:00
Charles Packer
264171f327 fix: patch streaming hidden reasoning event [LET-4167] (#4367)
* fix: patch streaming hidden reasoning event

* fix: patch reasoning_effort not getting passed to openai
2025-09-02 16:21:18 -07:00
Shubham Naik
98e8fcdee6 feat: allow list tools by tool type [PRO-870] (#4036)
* feat: allow list tools by tool type

* chore: update list

* chore: respond to comments

* chore: refactor tools hella

* Add tests to managers

* chore: branch

---------

Co-authored-by: Shubham Naik <shub@memgpt.ai>
Co-authored-by: Matt Zhou <mattzh1314@gmail.com>
2025-09-02 15:53:36 -07:00
Matthew Zhou
c69fc95870 feat: Extend crud lifecycle of messages [LET-4158] (#4364)
Extend crud lifecycle of messages
2025-09-02 14:38:30 -07:00
Matthew Zhou
8f425aa024 feat: Modify conversation search tool to be hybrid (#4362)
* Modify conversation search functionality

* Gate the roles
2025-09-02 13:45:53 -07:00
Matthew Zhou
6b625cb039 feat: Extract more text from messages [LET-4154] (#4359)
* Extract more message types

* Add extra tests testing for multiple parts
2025-09-02 13:25:32 -07:00
Charles Packer
12d39c083e fix: change the default reasoning effort for gpt-5 to be minimal, due to extremely high latency on medium (#4356)
fix: change the default reasoning effort for gpt-5 to be minimal, due to extremely high latency on it
2025-09-02 13:23:13 -07:00
Matthew Zhou
29502ac3aa feat: Add association for the vector namespace [LET-4148] (#4354)
Add association for the namespace
2025-09-02 13:21:42 -07:00
Charles Packer
9d49eff204 fix: patch the streaming issue in the openai client for when inner_thoughts_in_kwargs is off [LET-4146] (#4350)
fix: patch the streaming issue in the openai client for when inner_thoughts_in_kwargs is off
2025-09-02 12:44:25 -07:00
Matthew Zhou
b8a198688f feat: Embed all messages in turbopuffer [LET-4143] (#4352)
* wip

* Finish embedding

* Fix ruff and tests
2025-09-02 12:43:48 -07:00
Charles Packer
a696d9e3d5 fix: various fixes to make gpt-5 work better [LET-4138] (#4344)
* fix: patch gpt-5 compat

* feat: add verbosity level dropdown selector to llm config panel in ade

* fix: patch duplicated logic

* fix: make default verbosity None, just stage just publish

* fix: missing files
2025-09-02 10:30:25 -07:00
cthomas
9b315e9f54 feat: support filtering out approval messages for anthropic api (#4342) 2025-09-01 21:39:50 -07:00
Charles Packer
e741f84add fix: patch bug w/ extended thinking mode involving text leaking into reasoning (#4341)
* fix: patch for bad native reasoning behavior w/ sonnet

* fix: cleanup

* fix: cleanup

* fix: another prompt tune for less flaking
2025-09-01 20:26:24 -07:00
cthomas
f3df0433ae feat: add more tests for hitl (#4339) 2025-09-01 16:23:06 -07:00
Charles Packer
71a5eaa262 fix(core): change the backend mid-stream error packing to match what the FE expects [PRO-1107] (#4340)
fix(core): change the backend mid-stream error packing to match what the FE expects
2025-09-01 14:59:42 -07:00
cthomas
315d07568b feat: handle message persistence for approvals flows (#4338) 2025-09-01 14:10:02 -07:00
cthomas
1edcc13778 feat: support filtering out messages when converting to openai dict (#4337)
* feat: support filtering out messages when converting to openai dict

* fix imports
2025-09-01 12:48:45 -07:00
cthomas
2dcbb9d9a6 feat: add human in the loop tests to CI (#4335)
feat: add human in the loop tests-to-ci
2025-09-01 11:10:31 -07:00
Charles Packer
904bd2ba54 fix(core): init attempt at patching the extended thinking bug (#4331)
* fix(core): init attempt at patching the extended thinking bug

* fix: refactor, fix in a different way be merging instead of dropping, seems to be working

* chore: cleanup

* fix: patch the test - with the new native thinking behavior, anthropic native reasoning does not trigger properly post tool-call

* fix: another attempt

* fix: tests passing locally

* fix: another patch to tests

* fix: pathc more tests

* fix: try turning on the interleave-thinking beta header

* fix: patch test

* fix: add new tests to be more robust

* fix: patch the prompt, now it works more reliably

* fix: typo
2025-09-01 11:09:02 -07:00
cthomas
8d1389146a feat: support deny tool call request (#4334) 2025-09-01 10:33:18 -07:00
jnjpng
c9c9e727b8 fix: retry on MALFORMED_FUNCTION_CALL for gemini [LET-4089]
---------

Co-authored-by: Letta Bot <noreply@letta.com>
2025-09-01 07:26:13 -07:00
cthomas
86b073d726 feat: support approve tool call request (#4333) 2025-08-31 11:26:21 -07:00
Matthew Zhou
0854ba0d01 feat: Support timestamp filtering for archival memories [LET-3469] (#4330)
Finish temporal filtering
2025-08-30 19:31:07 -07:00
cthomas
6f9fa6ae44 feat: add input validation for messages field (#4328) 2025-08-29 22:06:12 -07:00
cthomas
9d03869287 feat: add error handling for approval response attempt (#4326)
* feat: add error handling for approval response attempt

* add one more error case

* improve error messages
2025-08-29 20:39:28 -07:00