issues Search Results · language:Dune language:Python language:JavaScript language:Java language:JavaScript language:Java
Filter by
55.2M results
Priority
P1
Problem
The project now spans chat, responses, tools, audio, model lifecycle, and local downloads. Compatibility can regress
unless request/response shapes are pinned with golden tests. ...
area:openai-compat
area:testing
priority:P1
Priority
P1
Problem
OpenAI-compatible clients commonly rely on /v1/embeddings. Gemma should not be forced into an embedding role; the server
needs a dedicated local embedding backend.
Scope
- Add ...
area:embeddings
area:openai-compat
priority:P1
Priority
P1
Problem
The transcription endpoints support core Whisper behavior, but OpenAI clients may request segment-level or word-level
timestamps through timestamp_granularities[].
Scope
- Parse ...
area:openai-compat
area:voice
priority:P1
Goal
Make aippocampus doctor provider --json expose only one action vocabulary in compact/default JSON: foreground-action-v1.
User-Visible Problem
Provider doctor compact JSON now has proper foreground_action ...
Severity: Nice to Have
The architecture specifies WAL mode with retry for concurrent access (Hermes sidecar, decay evaluator, CLI), but there
are no tests exercising concurrent writers. Should validate ...
Priority
P1
Problem
Gemma tool syntax is not one-to-one with OpenAI function calling. The current translation needs more edge-case coverage
for production clients.
Scope
- Expand parser support ...
area:openai-compat
area:tools
priority:P1
Severity: Nice to Have File: src/hpm/db.py
The merge_or_insert function either merges (updates existing) or inserts new. In neither case does it set superseded_by
on an existing entry. The column appears ...
Goal
Make bare aippocampus import --json behave as either a successful chooser or a failed needs-input recovery, and keep
write import paths out of the default safe action lane.
User-Visible Problem ...
Part of: #1169
Story: Fix Temporal Query Path — Inject Cache + Shared Executor
Background
The temporal query path in multi_search_service.py:_search_temporal_sync() constructs a FilesystemVectorStore ...
backlog
feat:temporal-performance
priority-1
story
Priority
P1
Problem
The Responses API shim needs stronger lifecycle behavior so clients relying on OpenAI response state can work
predictably.
Scope
- Add configurable local response persistence. ...
area:openai-compat
area:responses
priority:P1

Learn how you can use GitHub Issues to plan and track your work.
Save views for sprints, backlogs, teams, or releases. Rank, sort, and filter issues to suit the occasion. The possibilities are endless.Learn more about GitHub IssuesProTip! Restrict your search to the title by using the in:title qualifier.