Skip to content

docs: mark shipped roadmap items (#85, #50) #101

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
adaamko merged 1 commit into from
Sep 7, 2026
Merged

docs: mark shipped roadmap items (#85, #50) #101

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
6 changes: 2 additions & 4 deletions PUBLIC_ROADMAP.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -11,19 +11,17 @@ Open:
- [Real batched inference](https://github.com/KRLabsOrg/LettuceDetect/issues/23): `predict_prompt_batch` currently loops one sample at a time; the plan is padded-batch tokenization with a single forward pass.
- [Lighter installs](https://github.com/KRLabsOrg/LettuceDetect/issues/69): lazy top-level imports so the LLM detector never touches torch. How to slim the default install without breaking existing users is an open design question; input welcome on the issue.
- [Define evidence aggregation for chunked inference](https://github.com/KRLabsOrg/LettuceDetect/issues/74): benchmark whether evidence in any chunk or every chunk is required before changing the current conservative behavior.
- [Map repeated LLM span strings to distinct answer occurrences](https://github.com/KRLabsOrg/LettuceDetect/issues/85)

Shipped in this milestone so far: the network-free unit suite as a merge gate ([#73](https://github.com/KRLabsOrg/LettuceDetect/issues/73)), the `lettucedetect` CLI ([#47](https://github.com/KRLabsOrg/LettuceDetect/issues/47)), the detector latency/throughput benchmark ([#58](https://github.com/KRLabsOrg/LettuceDetect/issues/58)), token output from the LLM detector ([#65](https://github.com/KRLabsOrg/LettuceDetect/issues/65)), and the `min_confidence` documentation ([#64](https://github.com/KRLabsOrg/LettuceDetect/issues/64)).
Shipped in this milestone so far: the network-free unit suite as a merge gate ([#73](https://github.com/KRLabsOrg/LettuceDetect/issues/73)), the `lettucedetect` CLI ([#47](https://github.com/KRLabsOrg/LettuceDetect/issues/47)), the detector latency/throughput benchmark ([#58](https://github.com/KRLabsOrg/LettuceDetect/issues/58)), token output from the LLM detector ([#65](https://github.com/KRLabsOrg/LettuceDetect/issues/65)), the `min_confidence` documentation ([#64](https://github.com/KRLabsOrg/LettuceDetect/issues/64)), and deterministic localization of repeated LLM spans ([#85](https://github.com/KRLabsOrg/LettuceDetect/issues/85)).

## Next: [Validation — Typed-span workflows](https://github.com/KRLabsOrg/LettuceDetect/milestone/3)

The detectors already emit typed spans (category and subcategory per detected span). This milestone tests whether that localization improves real debugging, evaluation, monitoring, and agent workflows.

- [Dataset-level hallucination-rate evaluation and reporting](https://github.com/KRLabsOrg/LettuceDetect/issues/56)
- [A Claude Code hook that flags hallucinations in agent answers](https://github.com/KRLabsOrg/LettuceDetect/issues/50)
- [A stable HTTP contract for typed spans and detector selection](https://github.com/KRLabsOrg/LettuceDetect/issues/75)

Shipped: typed spans in the Streamlit demo ([#57](https://github.com/KRLabsOrg/LettuceDetect/issues/57)).
Shipped: typed spans in the Streamlit demo ([#57](https://github.com/KRLabsOrg/LettuceDetect/issues/57)) and the Claude Code Stop hook ([#50](https://github.com/KRLabsOrg/LettuceDetect/issues/50)).

## Exploring: [transferable span detection](https://github.com/KRLabsOrg/LettuceDetect/milestone/2)

Expand Down
Loading