-
Notifications
You must be signed in to change notification settings - Fork 3
All issues
Issue creation is restricted in this repository
Issues
is:issue state:open
is:issue state:open
Search results
calibrate, taxonomy and label plan must refuse on a judgeless task class
agent:readyNo human needed; an agent can finish this aloneNo human needed; an agent can finish this alonearea:clisrc/langchef/cli and its testssrc/langchef/cli and its testsgood first issueGood for newcomersGood for newcomerslifecycle:readySpecified well enough to start coldSpecified well enough to start coldpriority:P2When it comes upWhen it comes uptype:featureNew capabilityNew capabilityStatus: Open.#69 In deepskandpal/LangChef;[epic] Field trial: point it at a real application, offline then live
agent:needs-humanNeeds an account, a key, money, or a judgement callNeeds an account, a key, money, or a judgement callepicA container issue; the work is in its childrenA container issue; the work is in its childrenlifecycle:specNeeds a decision or a design before any codeNeeds a decision or a design before any codepriority:P1SoonSoontype:featureNew capabilityNew capabilityStatus: Open.#61 In deepskandpal/LangChef;Decision: what one label budget buys, estimation or diagnosis
area:coresrc/langchef/core and its testssrc/langchef/core and its testslifecycle:specNeeds a decision or a design before any codeNeeds a decision or a design before any codepriority:P1SoonSoontype:decisionA call to make and record in DECISIONS.mdA call to make and record in DECISIONS.mdStatus: Open.#60 In deepskandpal/LangChef;The remaining regression knobs from the build order
agent:readyNo human needed; an agent can finish this aloneNo human needed; an agent can finish this alonearea:dogfoodsrc/langchef/dogfood and its testssrc/langchef/dogfood and its testslifecycle:readySpecified well enough to start coldSpecified well enough to start coldpriority:P2When it comes upWhen it comes upsize:MTwo to eight hoursTwo to eight hourstype:infraCI, toolchain, packaging, releaseCI, toolchain, packaging, releaseStatus: Open.The self-test nobody else can run: an ROC curve for the detector itself
agent:readyNo human needed; an agent can finish this aloneNo human needed; an agent can finish this alonearea:dogfoodsrc/langchef/dogfood and its testssrc/langchef/dogfood and its testslifecycle:specNeeds a decision or a design before any codeNeeds a decision or a design before any codepriority:P1SoonSoonsize:LMore than eight hoursMore than eight hourstype:researchAnswering a question, not buildingAnswering a question, not buildingStatus: Open.Cost and latency reported alongside quality, always
agent:readyNo human needed; an agent can finish this aloneNo human needed; an agent can finish this alonearea:coresrc/langchef/core and its testssrc/langchef/core and its testslifecycle:specNeeds a decision or a design before any codeNeeds a decision or a design before any codepriority:P1SoonSoonsize:MTwo to eight hoursTwo to eight hourstype:featureNew capabilityNew capabilityStatus: Open.Golden-set maintenance: mine failures, retire stale cases with a stated reason
agent:readyNo human needed; an agent can finish this aloneNo human needed; an agent can finish this alonearea:workspacesrc/langchef/workspace and its testssrc/langchef/workspace and its testslifecycle:specNeeds a decision or a design before any codeNeeds a decision or a design before any codepriority:P2When it comes upWhen it comes upsize:LMore than eight hoursMore than eight hourstype:featureNew capabilityNew capabilityStatus: Open.Regression thresholds derived from observed variance, not guessed constants
agent:readyNo human needed; an agent can finish this aloneNo human needed; an agent can finish this alonearea:coresrc/langchef/core and its testssrc/langchef/core and its testslifecycle:specNeeds a decision or a design before any codeNeeds a decision or a design before any codepriority:P1SoonSoonsize:MTwo to eight hoursTwo to eight hourstype:featureNew capabilityNew capabilityStatus: Open.Delivery into Slack or the issue tracker as content, not a pointer
agent:needs-humanNeeds an account, a key, money, or a judgement callNeeds an account, a key, money, or a judgement callarea:adapterssrc/langchef/adapters and its testssrc/langchef/adapters and its testslifecycle:specNeeds a decision or a design before any codeNeeds a decision or a design before any codepriority:P2When it comes upWhen it comes upsize:MTwo to eight hoursTwo to eight hourstype:featureNew capabilityNew capabilityStatus: Open.Weekly recalibration sized to a ten-minute labelling budget
agent:readyNo human needed; an agent can finish this aloneNo human needed; an agent can finish this alonearea:coresrc/langchef/core and its testssrc/langchef/core and its testslifecycle:specNeeds a decision or a design before any codeNeeds a decision or a design before any codepriority:P1SoonSoonsize:MTwo to eight hoursTwo to eight hourstype:featureNew capabilityNew capabilityStatus: Open.Scheduling through cron, CI, or the harness scheduler
agent:readyNo human needed; an agent can finish this aloneNo human needed; an agent can finish this alonearea:adapterssrc/langchef/adapters and its testssrc/langchef/adapters and its testslifecycle:specNeeds a decision or a design before any codeNeeds a decision or a design before any codepriority:P1SoonSoonsize:MTwo to eight hoursTwo to eight hourstype:featureNew capabilityNew capabilityStatus: Open.sampleis contracted but unbuilt, andconnect/is emptyagent:readyNo human needed; an agent can finish this aloneNo human needed; an agent can finish this alonearea:connectsrc/langchef/connect and its testssrc/langchef/connect and its testslifecycle:specNeeds a decision or a design before any codeNeeds a decision or a design before any codepriority:P2When it comes upWhen it comes upsize:LMore than eight hoursMore than eight hourstype:featureNew capabilityNew capabilityStatus: Open.