fbf003d7ec
Add comment
2025-07-25 06:32:40 +02:00
8e795f183d
Update system message to highlight verbs
2025-07-24 13:42:19 -05:00
18597aabad
Make sure we sort closed cards by recency by default
...
So that the filter used for the insight context fetches the most recent cards when using indexed_by closed
2025-07-24 17:46:40 +02:00
95e065a0c4
Reduce the injected load a bit
2025-07-24 17:33:40 +02:00
523f3f0acb
Set some hard limits on the prompt lenghts we inject
2025-07-24 17:32:50 +02:00
0429dc28f3
Fix for injecting the current context
...
This is a workaround to make the prototype work. We create a context on the fly to execute AI-derived
commands. When those require listing cards, we ended up losing the notion of "being on a perma screen".
2025-07-24 17:27:32 +02:00
b2ee79bbd0
Format
2025-07-24 16:48:06 +02:00
e61ee25b47
Suite passing
2025-07-24 16:48:05 +02:00
4574c0d85c
Include stage information about the card
2025-07-24 16:48:05 +02:00
ba97ec05d6
Remove transaction as the AI commands can hold the transaction for too long
2025-07-24 16:48:05 +02:00
5ad2d09eae
Move the hack to fetch completed cards to the insight command logic
...
We need to rework this as part of the filters revamp. "indexed_by" being
a mix of filters keeps biting us.
2025-07-24 16:48:05 +02:00
fc26329daf
Restore default to open cards. Use special "all" value instead.
...
The change could have side effects (e.g: when running commands).
This also include further changes.
2025-07-24 16:48:04 +02:00
018c746d2a
Make the prompt smarter about the current context to make better calls
2025-07-24 16:48:03 +02:00
8bb7783215
Extract "prompt generation" logic to models for events, comments and cards
...
So that we can reuse in other scenarios
2025-07-24 16:48:03 +02:00
dd28e3d1e9
Replace special value with just not asuming that filters are about open cards by default
2025-07-24 16:48:03 +02:00
9d0887f6bd
Use a workaround with a special indexed_by filter to "query all"
2025-07-24 16:48:02 +02:00
83d743fe2a
Make the translator use the /insight command
2025-07-24 16:48:02 +02:00
7a8bf4895c
More prompt fixes
2025-07-24 16:48:01 +02:00
4d3e7bf0af
Fix: llm model
2025-07-24 16:48:01 +02:00
1c6adf54a8
Restore semantic searches
2025-07-24 16:47:57 +02:00
df9c1e0674
Make sure we pass expressions as a single term
...
Multiple terms refine the search on each term and works way
worse with the semantic approach
2025-07-24 16:47:54 +02:00
6080c2aa9b
Refine when to filter by cards, make it more relaxes with the expressions it accepts
2025-07-24 16:47:53 +02:00
4276aad5a7
Mark notifications as unread from the index
2025-07-23 17:01:59 -05:00
dfb6198e94
Merge branch 'main' into web-push
...
* main: (50 commits)
Regenerate fixtures
Let summaries know about collections
Regenerate fixtures
More tweaks
Regenerate fixtures after tweaking prompt
Fix model selection (I had broken it!)
Some prompt refinements
Some of us have basecamp.com email addresses
Update nokogiri and thor to addres CVEs
Fix tests
Format
Move model options to helper
Tidy up the admin tool for prompts
Extract methods
Format
Default to the current summary
Make models selectable
Add an animated effect for the generating state
Remove debug
Wrap them both in the container so the generating state looks right
...
2025-07-23 11:50:35 -05:00
f3fa324f09
Let summaries know about collections
2025-07-23 16:54:39 +02:00
c082a1b1f5
More tweaks
2025-07-23 13:15:49 +02:00
c8c297a75e
Fix model selection (I had broken it!)
2025-07-23 13:10:47 +02:00
49518692bb
Some prompt refinements
2025-07-23 13:04:41 +02:00
94ba4f6074
Some of us have basecamp.com email addresses
2025-07-23 12:50:07 +02:00
fc1ff7bf5e
Fix tests
2025-07-23 12:21:20 +02:00
90ddadea5c
Format
2025-07-23 12:04:18 +02:00
c0b385b19b
Tidy up the admin tool for prompts
2025-07-23 11:55:44 +02:00
f32e89c277
Format
2025-07-23 11:43:10 +02:00
c3ae05833e
Make models selectable
2025-07-23 11:02:34 +02:00
f413e19508
Prefer Account.sole
2025-07-22 22:45:23 -05:00
5322ce2afb
Add comment's context for each card too
2025-07-22 22:43:01 +02:00
e1ab4ec13c
Use 4o for now, it's just better for summarizing and conversation
...
It's more expensive, going with the best and we'll tune later
2025-07-22 22:34:12 +02:00
662f6dffe8
Use gpt-4.1 for now, several prompt refinements
2025-07-22 21:32:10 +02:00
fd0484176c
Format
2025-07-22 18:59:10 +02:00
e87ed26f75
Add tests
2025-07-22 18:45:07 +02:00
6aca86c87a
Forgot to rename this
2025-07-22 18:31:52 +02:00
13b20d43c8
Refine prompt
2025-07-22 18:13:44 +02:00
ca640b5b45
Remove (not used)
2025-07-22 15:39:40 +02:00
26089f77b7
Remove unused column and rename contents => content
2025-07-22 15:23:39 +02:00
8f6e81a687
Only summarize today if there are at least 10 events
...
https://3.basecamp.com/2914079/buckets/37331921/todos/8877635813#__recording_8879434338
2025-07-22 15:04:55 +02:00
6ba1884fb0
Fix: summaries where failing to generate with filters
2025-07-22 14:59:38 +02:00
0397327442
Avoid links to comments for now
2025-07-22 14:40:59 +02:00
6d26195bc7
Add a debug prompt endpoint so that we can play with the prompts live
2025-07-22 14:33:34 +02:00
89a9e7f32c
Tweak prompt for trying to avoid liks like "card 1234"
2025-07-22 12:28:05 +02:00
9c6f1354c4
Broadcast and show summaries on the fly
2025-07-22 12:26:32 +02:00
Jorge Manrubia