storied
A 5e storytelling engine with an LLM DM
Star
Fork
Watch

Commits on python-and-claude
Move to Opus 5 and keep input history with the player
committed
Heal list-shaped skills/tools and keep wikilinks out of narration
committed
Tighten quality gates and refactor the test_execute_tool fixtures
committed
Forge per-campaign cultures and use them to name everything
committed
Collapse ToolContext into a real singleton
committed
Plot a campaign arc before seeding the world
committed
Fixing onboarding stuff
committed
Move campaign data to ~/.storied and add a user homebrew layer
committed
Make the character toolkit dumber so the DM runs the rules
committed
Lean on the tools harder for 5e bookkeeping
committed
Nudge the DM when advancement has been pending for a while
committed
Move the tool surface to FastMCP and tighten the input schemas
committed
Rewrite character sheet as structured data with bookkeeping tools
committed
Add adjust_coins tool and purse clamping
committed
Show advancement_ready flag in DM character context
committed
Add background advancement evaluator and DM notification channel
committed
Fix /context to show all sections and use a 40x5 grid
committed
Flush renderer word buffer when deferred tools start
committed
Fix roll() signature mismatch in sandbox and clean up run_code docs
committed
Add sandboxed code execution for the DM via Pydantic Monty
committed
Clean up spacing between narrative text and tool notifications
committed
Add word wrapping to the stream renderer
committed
Add initiative tracking tools for structured combat
committed
Replace Rich Live/Markdown with streaming ANSI renderer
committed
Split equipment table rows and add engine tests
committed
Fix SRD chunking so recall actually finds class features and conditions
committed
Add self-tunable DM style via tune tool
committed
Add local semantic vector search for recall
committed
Add running game transcript so the DM remembers what was said
committed
Enable bracketed paste so multi-line pastes don't flood the DM
committed
Add entity path index and write-through cache for faster tool calls
committed
Add display blocks, sandbox mode, column layout, and test coverage
committed
Replace LLM-formatted /status and /me with deterministic rendering
committed
Consolidate claude subprocess calls, add currency denominations, polish UI
committed
Fix DM context loss on resume, add world tick and toolkit improvements
committed
Merge mark_time into set_scene, fix day-roll data loss
committed
Switch from Anthropic SDK to claude -p with in-process MCP
committed
Add world seeding, planner, and new slash commands
committed
Switch to Haiku 4.5, fix cost tracking, add slash commands
committed
Add entity model, transcript logging, and HP clamping
committed
Refactor world state architecture and tool semantics
committed
Use STORIED_ANTHROPIC_API_KEY to avoid collision with Claude Code
committed
Add prompt caching for ~80% input token savings
committed
Add character creation, session management, and DM improvements
committed
Add session state, campaign log, and character persistence
committed
Add DM engine with dice rolling and rule lookups
committed
Initial project setup with SRD extraction pipeline
committed