Issues
- 1
JSON: convert oneOf -> anyOf when safe
#77 opened by mmoskal - 1
slow JSON schema -> llg conversion
#98 opened by mmoskal - 6
compile enums as regexes, not grammars
#84 opened by mmoskal - 1
missing `"` in output for string schema
#81 opened by mmoskal - 5
- 0
implement `additionalItems`
#86 opened by mmoskal - 0
- 2
we only bail on unknown formats for strings
#85 opened by mmoskal - 0
Json regexes should not be anchored by default
#66 opened by hudson-ai - 5
Grammar rules should be unique
#58 opened by v-jkegler - 1
document JSON schema support
#76 opened by mmoskal - 0
Change `process_prompt` api to return a (backtrack, ff_tokens) tuple consistent with `commit_token`
#69 opened by hudson-ai - 0
Make `process_prompt` respect backtracking/fast-forwarding inference capabilities
#68 opened by hudson-ai - 0
[perf] large `allowed_lexemes` sets
#64 opened by mmoskal - 1
fix advance_parser()
#8 opened by mmoskal - 2
greedy lexer definition
#57 opened by mmoskal - 5
Install of llguidance on Pi 5
#23 opened by v-jkegler - 8
- 0
- 0
add API to "fast-forward" grammar
#47 opened by mmoskal - 0
support non-contextual keywords
#46 opened by mmoskal - 0
only allowing valid tokenizations in grammars
#1 opened by mmoskal - 0
- 11
Nullables bug
#11 opened by v-jkegler - 3
- 3
Tokenization inconsistency between prompt and fast-forward at beginning of grammar
#4 opened by hudson-ai - 0
- 1
- 3
`usize` overflow in `num_tokens` when completing a `Gen` with nonempty `stop_regex`
#6 opened by hudson-ai - 0
LLGuidance should emit warnings/logs when lexemes/subgrammars are cut short by max_tokens
#5 opened by hudson-ai - 0