AK-Bench
benchmark report
dashboard
agents
runtimes
plugins
models
tasks
runs
docs
Documentation
/
Spike
Documentation overview
Project
1
AK-Bench
Handbook
18
Agent adapters
Agent integration — spike results (Phase 0)
CLI
Datasets, tasks and workspaces (Phase 3)
Descriptor system (Phase 2)
Development environment
Evaluator Engine
Experiments and planning
Export
Knowledge Replay
Metrics and Observability
Plugin Configuration and Observability
Reporting
Reproducibility
LLM Router integration — spike results (Phase 0)
Runtime Manager
Run Orchestrator
Storage
Datasets
31
Readme
bookshop
cadence
Depweaver
Resolution and repair contract
freight
ledgerlite
notesapi
stockpile
Streamforge
textkit
Typeweave
Type inference implementation contract
WALQueue
Cancelling a confirmed order leaves the stock reserved
The five percent tier is lost at exactly one hundred dollars
A worker with a stale epoch can take a live lease back
Repair the resolver's branch and context protocol
A conversion answers with the oldest rate in force
A write-off destroys the reserved part of an expired lot
Repair durable queue recovery
Word counting must ignore repeated whitespace
Implement `plan_catch_up` from its specification
Implement `shop` from its specification
Implement `reconcile` from its specification
Implement `search_notes` from its specification
Implement `merge` from its specification
Implement `slugify` from the specification
Implement `allocate` from its specification
Implement a restartable event-time engine
Complete the type checker of an existing compiler
Plan
1
AK-Bench --- kompletny plan implementacji
Spike
2
Phase 0 — integration spike
Spike fixture repo
spike/repo/README.md
Spike fixture repo
Small repository used by the AK-Bench phase 0 spike.