@percepta/kaizen 0.6.0 → 0.7.0

package/README.md

@@ -1,176 +1,104 @@
 # Kaizen
 
-Automated AI researcher that improves AI systems. Kaizen investigates production traces, builds evaluation datasets, records scored runs, and helps prepare improvements -- driven by Claude Code with a live dashboard.
+Kaizen is an agentic eval platform for AI systems. It helps a coding agent create a system definition, curate a Langfuse-backed dataset, write an eval script, run a baseline, and iterate on variants while Kaizen records scored runs under `kaizen/.kaizen/runs/`.
 
 ## Install In A Target Repo
 
-For a persistent local `kaizen` command:
-
 ```bash
 npm install -g @percepta/kaizen
 kaizen init
-kaizen create system <system-id> # add --eval-language ts for TypeScript
-kaizen run --system <system-id> --variant baseline --hypothesis "starting baseline"
+kaizen guide
+kaizen create system <system-id>
+kaizen create view <system-id> --type trace
+kaizen create view <system-id> --type dataset-item
+kaizen run --system <system-id> --variant baseline --diagnostic --hypothesis "starting baseline"
 kaizen studio
 ```
 
-For one-off use without a global install:
+For one-off use:
 
 ```bash
 npx @percepta/kaizen init
 ```
 
-For a repo-local dev dependency:
+Kaizen is installed inside the customer repo. The customer-owned footprint is intentionally small:
 
-```bash
-pnpm add -D @percepta/kaizen
-pnpm exec kaizen studio
-```
+- `kaizen/config.ts`
+- `kaizen/systems/<system-id>/system.md`
+- `kaizen/systems/<system-id>/eval.py|ts`
+- optional `kaizen/systems/<system-id>/trace.tsx`
+- optional `kaizen/systems/<system-id>/dataset-item.tsx`
+- optional `kaizen/systems/<system-id>/rubric.md`
+- `kaizen/.kaizen/runs/`
 
-Kaizen is installed inside the customer repo. System definitions, eval scripts, custom views, and `.kaizen/runs/` all live there; the CLI, runner, dashboard shell, and recipes come from the package.
-
-## Developing This Repo
-
-```bash
-pnpm install
-pnpm --filter @percepta/kaizen dev:studio
-```
+Package-owned agent guidance is printed with `kaizen guide`. Customer-specific durable notes belong in `kaizen/systems/<system-id>/system.md`; Kaizen does not create repo-level agent markdown such as `KAIZEN.md`, `AGENTS.md`, or `CLAUDE.md`.
 
-This starts the dashboard at http://localhost:6789 against `examples/legacy-workspace`, a transitional fixture with historical customer and system definitions. The CLI and dashboard both live in `packages/kaizen` (CLI in `src/`, Next.js dashboard in `dashboard/`).
+## Lifecycle
 
-Other dev scripts:
+1. Run `kaizen init` once in the target repo.
+2. Run `kaizen create system <system-id>` and fill in `kaizen/systems/<system-id>/system.md`.
+3. Use Studio Data to create or select a Langfuse dataset, add useful source traces, and label dataset items.
+4. Replace `kaizen/systems/<system-id>/eval.py|ts` with a real eval that reads the dataset named by `dataset_version`.
+5. Run a diagnostic baseline, then a full baseline.
+6. Run variants with `kaizen run`, inspect `kaizen log`, and use Studio to compare runs and failures.
 
-| Script                                      | What it does                      |
-| ------------------------------------------- | --------------------------------- |
-| `pnpm --filter @percepta/kaizen dev:studio` | Start the Studio dashboard        |
-| `pnpm --filter @percepta/kaizen dev:next`   | Start only the Next.js dev server |
-| `pnpm typecheck`                            | Typecheck all packages            |
-| `pnpm test`                                 | Run package tests                 |
+The eval script emits NDJSON events to `--out-fd`; the runner owns process supervision, `kaizen/.kaizen/runs/`, crash recording, and automatic promotion. For Langfuse-backed evals, the eval should also link each dataset item to the fresh trace generated by that run and write the primary metric as a trace score.
 
-## Publishing
+## Custom Views
 
-Publishing `@percepta/kaizen` to npm is automated with Changesets. For changes
-that affect the published package, add a changeset:
+Custom views are plain React components co-located with the system:
 
 ```bash
-pnpm changeset
+kaizen create view <system-id> --type trace
+kaizen create view <system-id> --type dataset-item
 ```
 
-Merging to `main` runs `.github/workflows/build-and-publish.yml`, which builds
-the CLI and bundled Studio, then either opens a version PR or publishes to npm
-using the `NPM_TOKEN` repository secret.
-
-## How It Works
-
-Kaizen closes the eval loop for AI systems:
-
-1. **Investigate** -- pull production traces from Langfuse, analyze failure patterns
-2. **Build dataset** -- create versioned eval datasets from traces with ground truth
-3. **Annotate** -- label ground truth via the dashboard's inline annotation view
-4. **Record runs** -- test system variants against ground truth, scored automatically
-5. **Improve** -- prepare a PR from the latest promoted baseline when a human asks
-
-The `/kaizen` slash command in Claude Code orchestrates this workflow. Variant-builder agents can execute in parallel worktrees, but they pass the main checkout's `.kaizen` path via `KAIZEN_STATE_DIR` or `--state-dir` so the dashboard always reads one canonical state tree.
-
-## Dashboard
-
-The web app (Next.js, pages router) provides:
+`trace.tsx` receives the full Langfuse trace payload plus actions for writing scores. `dataset-item.tsx` receives the dataset item, the linked source trace when available, and actions for updating the dataset item or linking run items. Browser-side credentials are not required; Studio proxies the write actions through local API routes.
 
-- **Data** -- inspect Langfuse datasets, dataset items, and source traces
-- **Experiments** -- inspect local Kaizen runs from the customer repo's `.kaizen/runs/` store
-- **Ideas** -- inspect Linear issues scoped to the system's configured project and the shared `Kaizen` label
-- **Source indicators** -- show whether a field is sourced from repo code, Langfuse, Linear, or the local filesystem
+Run `kaizen guide views` for the exact prop and action interfaces.
 
-### Keyboard Shortcuts
-
-| Shortcut | Action         |
-| -------- | -------------- |
-| `Cmd+[`  | Toggle sidebar |
-| `Cmd+/`  | Show shortcuts |
-
-## System Definitions
-
-In real use, each target/customer repo owns its own `customers/`, `systems/`, `rubrics/`, `eval/`, and optional `views/` directories. Each system is defined in `systems/*.md` with YAML frontmatter:
+## Developing This Repo
 
-```yaml
-run_eval: eval/<system>.ts # or .py
-eval_version: 1
-dataset_version: v1
-eval_style: ground-truth
-primary_metric: score
+```bash
+pnpm install
+pnpm --filter @percepta/kaizen dev:studio
 ```
 
-The eval script emits NDJSON events to `--out-fd`; the runner owns `.kaizen/runs/`.
-Kaizen runs Python evals with `python3`, JavaScript evals with `node`, and
-TypeScript evals with the package's bundled `tsx` loader. New system scaffolds
-default to Python; pass `--eval-language ts` to create a TypeScript eval.
-For Langfuse-backed production evals, the same script should also link each
-dataset item to the fresh trace produced by that run in a Langfuse dataset run
-and write the primary metric as a trace score. Those writes are for durable
-trace inspection; the NDJSON `complete.score` remains Kaizen's required result
-contract.
-
-This repo keeps historical definitions under `examples/legacy-workspace/` only as sample data for local Studio development:
-
-| Customer         | System                       | Primary Metric          |
-| ---------------- | ---------------------------- | ----------------------- |
-| Transcarent      | EMO HIE Processing           | F2                      |
-| Transcarent      | EMO Facility Processing      | F2                      |
-| Transcarent      | EMO Cost Savings Agent       | Classification Accuracy |
-| Transcarent      | EMO Summarization            | --                      |
-| Transcarent      | Orbit Call Summarization     | Judge Quality           |
-| Cityblock Health | BOI Chaselist Impact         | Calibration Error       |
-| Cityblock Health | Concurrent Review Agent      | --                      |
-| Cityblock Health | Contract Exclusion Detection | --                      |
-| Cityblock Health | Quality Gap Modeling         | Calibration Error       |
-| Janus Henderson  | Portfolio Analytics          | --                      |
-| Summa Health     | Agentic BI (SLCC)            | --                      |
-
-## Repository Structure
+This starts Studio at `http://localhost:6789` against `examples/demo-workspace`, a local fixture for package development. The CLI lives in `src/`; the bundled Next.js Studio lives in `dashboard/`.
 
-```
-kaizen/
-├── packages/kaizen/                 # Published @percepta/kaizen package
-│   ├── src/                         # CLI source
-│   ├── dashboard/                   # Next.js Studio (built into the published bundle)
-│   └── examples/legacy-workspace/   # Transitional customer/system fixture for local dev
-```
-
-## Tech Stack
+Useful scripts:
 
-- **Frontend**: Next.js (pages router), TypeScript, CSS modules, dark theme
-- **AI**: Claude Code with `/kaizen` skill + variant-builder agents
-- **Observability**: Langfuse for traces, datasets, and annotation state; `.kaizen/runs/` for local run truth
-- **Package manager**: pnpm
+| Script                                      | What it does                       |
+| ------------------------------------------- | ---------------------------------- |
+| `pnpm --filter @percepta/kaizen dev:studio` | Start Studio with the demo fixture |
+| `pnpm --filter @percepta/kaizen dev:next`   | Start only the Next.js dev server  |
+| `pnpm --filter @percepta/kaizen typecheck`  | Typecheck the package              |
+| `pnpm --filter @percepta/kaizen test`       | Run package tests                  |
 
-## Environment Setup
+## Environment
 
-Create `.env.local` in the workspace repo root with Kaizen credentials:
+Create `.env.local` in the workspace repo root:
 
-```
+```text
 LANGFUSE_HOST=https://...
 LANGFUSE_PUBLIC_KEY=pk-lf-...
 LANGFUSE_SECRET_KEY=sk-lf-...
 LINEAR_API_KEY=lin_api_...
+LINEAR_TEAM_KEY=ENG
 ```
 
-Kaizen Studio reads Langfuse and Linear credentials from the workspace root
-`.env.local`. Put the values there instead of app package env files so stale
-package-level placeholders cannot shadow the credentials Studio needs.
-
-Langfuse credentials power the Data surface. `LINEAR_API_KEY` powers the Ideas
-surface and the `kaizen ideas --system <id>` CLI command.
+Langfuse credentials power the Data surface and custom view actions. `LINEAR_API_KEY` and `LINEAR_TEAM_KEY` power `kaizen ideas --system <id>`.
 
-System Ideas configuration should use a stable Linear project URL or ID:
+System Ideas configuration should use a stable Linear project URL or ID in `system.md`:
 
 ```yaml
-linear_project: https://linear.app/aitco/project/kaizen-v0-555399b53e23
+linear_project: https://linear.app/<workspace>/project/<project-slug>
 ```
 
-Project names are intentionally not used for the connection because they can
-change in Linear without changing project identity.
+## Publishing
 
-## Docs
+Publishing `@percepta/kaizen` to npm is automated with Changesets. For changes that affect the published package, add a changeset:
 
-- [Langfuse Standards](docs/langfuse-standards.md) -- how we structure Langfuse across all customers
-- [Eval Framework](docs/eval-framework.md) -- evaluation philosophy and requirements
+```bash
+pnpm changeset
+```

package/agent/claude-command.md

@@ -0,0 +1,23 @@
+# Kaizen Claude Command Guide
+
+Use this as a Claude Code command body when Claude should drive the Kaizen lifecycle. It is package-owned; do not copy it into customer repos as durable markdown unless the user explicitly asks.
+
+## Rules
+
+- Never commit PHI or credentials.
+- Run `kaizen guide` first when guidance is not already in context.
+- Put customer-specific notes in `kaizen/systems/<system-id>/system.md`.
+- Use Studio for dataset curation and custom dataset item views for labeling workflows.
+
+## Workflow
+
+1. Select a system. If none is given, list `kaizen/systems/*/system.md` and ask which one.
+2. Read `system.md`, relevant application code, and current `kaizen log --system <system-id> --json`.
+3. If the system is new, run `kaizen create system <system-id> --eval-language py|ts` and fill in the scaffold.
+4. Use Studio Data to create/select a dataset, add traces, and label expected outputs.
+5. Replace the starter eval with real code that loads `--dataset`, runs the candidate, emits NDJSON events, and persists Langfuse links/scores when available.
+6. Run a diagnostic baseline, then the full baseline.
+7. Iterate variants with `kaizen run`.
+8. Create `trace.tsx` or `dataset-item.tsx` only when the default views are insufficient.
+
+For exact eval and view contracts, run `kaizen guide evals` and `kaizen guide views`.

package/agent/evals.md

@@ -0,0 +1,41 @@
+# Kaizen Eval Guide
+
+Eval scripts are customer-owned executable code stored at `kaizen/systems/<system-id>/eval.py|ts`. `kaizen run` invokes the path named by `run_eval` in `kaizen/systems/<system-id>/system.md`:
+
+```bash
+<run_eval> --variant <variant-id> --dataset <dataset_version> --out-fd 3 [--max-items <n>]
+```
+
+The eval must write NDJSON events to `--out-fd`. Do not write these events to normal stdout.
+
+```json
+{"type":"start","n":10,"eval_version":1,"dataset_version":"v1"}
+{"type":"item","id":"item-1","score":0.8,"breakdown":{"score":0.8},"trace_id":"trace-id-or-null"}
+{"type":"complete","score":0.82,"n":10,"breakdown":{"score":0.82},"worst_traces":[{"id":"item-1","score":0.8,"trace_id":"trace-id-or-null"}]}
+```
+
+The terminal `complete.score` is Kaizen's authoritative result. It must be a number in `[0, 1]`.
+
+## Langfuse Persistence
+
+For Langfuse-backed evals:
+
+- Treat `--dataset` as the Langfuse dataset name unless `system.md` says otherwise.
+- Load dataset items from that dataset.
+- Run the candidate system for each item.
+- Capture the fresh Langfuse trace id for that item.
+- Link the dataset item to the fresh trace in a Langfuse dataset run.
+- Write the primary metric as a Langfuse score on the fresh trace.
+- Emit the same item score through Kaizen's NDJSON stream.
+
+Langfuse stores trace inspection, dataset-run history, and score metadata. `kaizen/.kaizen/runs/` remains the source of truth for promotion and run state.
+
+## Baseline
+
+Run a diagnostic baseline first:
+
+```bash
+kaizen run --system <system-id> --variant baseline --diagnostic --hypothesis "starting baseline"
+```
+
+If setup, credentials, dataset access, and event schema are valid, run the full baseline without `--diagnostic`.

package/agent/overview.md

@@ -0,0 +1,53 @@
+# Kaizen Agent Guide
+
+Kaizen helps a coding agent define, evaluate, inspect, and improve an AI system inside the customer repo. This guide is package-owned; rerun `kaizen guide` after package upgrades.
+
+Do not create extra long-lived agent markdown files. Customer-specific notes belong in `kaizen/systems/<system-id>/system.md`; repo-owned code belongs beside it in `kaizen/systems/<system-id>/`.
+
+## Commands
+
+Run commands from the repo root:
+
+- `kaizen init` - scaffold Kaizen once.
+- `kaizen guide topics` - list focused guide topics.
+- `kaizen create system <system-id> --eval-language py|ts` - create `kaizen/systems/<system-id>/system.md` and `kaizen/systems/<system-id>/eval.py|ts`.
+- `kaizen create view <system-id> --type trace` - create `kaizen/systems/<system-id>/trace.tsx`.
+- `kaizen create view <system-id> --type dataset-item` - create `kaizen/systems/<system-id>/dataset-item.tsx`.
+- `kaizen studio` - open Studio for dataset curation, trace inspection, and run review.
+- `kaizen run --system <system-id> --variant <variant-id> --hypothesis "<why>"` - record one eval run.
+- `kaizen run --system <system-id> --variant <variant-id> --diagnostic --hypothesis "<why>"` - run a small diagnostic sample first.
+- `kaizen log --system <system-id> --json` - inspect the promoted baseline and recent runs.
+
+Run state is written to `kaizen/.kaizen/`. When evaluating from a Git linked worktree, Kaizen automatically stores run state in the primary checkout's `kaizen/.kaizen/`.
+
+## Files
+
+- `kaizen/systems/<system-id>/system.md` is the durable system definition. It should explain the workflow, key files, setup, dataset, metric, known failures, and variant ideas.
+- `kaizen/systems/<system-id>/eval.py|ts` is the eval entrypoint named by `run_eval`.
+- `kaizen/systems/<system-id>/trace.tsx` is an optional custom trace view.
+- `kaizen/systems/<system-id>/dataset-item.tsx` is an optional custom dataset labeling view.
+- `kaizen/systems/<system-id>/rubric.md` is optional and only needed for LLM-as-judge or hybrid evals.
+
+Each `system.md` must include:
+
+```yaml
+run_eval: kaizen/systems/<system-id>/eval.py
+eval_version: 1
+dataset_version: <langfuse-dataset-name>
+eval_style: ground-truth
+primary_metric: score
+target: 0.90
+```
+
+## Lifecycle
+
+1. Run `kaizen create system <system-id>` unless the system already exists.
+2. Read the codebase and fill in `system.md` with real key files, setup, data sources, dataset, and metric.
+3. Use Studio Data to create/select a dataset, add representative traces, and label expected outputs.
+4. Replace the starter eval with real code that loads the dataset named by `--dataset`.
+5. Run a diagnostic baseline.
+6. Run the full baseline.
+7. Iterate on variants with `kaizen run`; read `kaizen log` and Studio failures between attempts.
+8. Create custom views only when the default JSON views are not enough for trace inspection or dataset labeling.
+
+For eval details, run `kaizen guide evals`. For view props and actions, run `kaizen guide views`.

package/agent/variant-builder.md

@@ -0,0 +1,22 @@
+# Kaizen Variant Builder Guide
+
+You implement and evaluate one variant, record one run with `kaizen run`, then stop.
+
+## Setup
+
+1. Work in the assigned worktree, not the main checkout.
+2. Let Kaizen auto-detect the primary checkout for run state. Runs from linked worktrees are recorded under the primary checkout's `kaizen/.kaizen/`.
+3. Read `kaizen/systems/<system-id>/system.md`, the parent run manifest when present, and the parent failures.
+4. Install or start only what the system setup section requires.
+
+## Run
+
+```bash
+kaizen run \
+  --system <system-id> \
+  --variant <variant-id> \
+  --parent <parent-run-id> \
+  --hypothesis "<what changed and why>"
+```
+
+The runner owns process supervision, `kaizen/.kaizen/runs/`, crash recording, and promotion. Read the single summary line it prints and include the run id and score in your handoff.

package/agent/views.md

@@ -0,0 +1,51 @@
+# Kaizen Custom Views Guide
+
+Custom views are customer-owned React components co-located with the system.
+
+```bash
+kaizen create view <system-id> --type trace
+kaizen create view <system-id> --type dataset-item
+```
+
+Studio loads:
+
+- `kaizen/systems/<system-id>/trace.tsx`
+- `kaizen/systems/<system-id>/dataset-item.tsx`
+
+No `system.md` frontmatter field is required.
+
+## Trace View
+
+```tsx
+import type { TraceRendererProps } from "@percepta/kaizen";
+
+export default function TraceView({ trace, actions }: TraceRendererProps) {
+  return <pre>{JSON.stringify(trace, null, 2)}</pre>;
+}
+```
+
+Trace views receive `{ trace, context, actions }`. `actions.createScore(...)` writes a Langfuse score for the current or supplied trace id.
+
+## Dataset Item View
+
+```tsx
+import type { DatasetItemRendererProps } from "@percepta/kaizen";
+
+export default function DatasetItemView({
+  datasetItem,
+  trace,
+  actions,
+}: DatasetItemRendererProps) {
+  return <pre>{JSON.stringify({ datasetItem, trace }, null, 2)}</pre>;
+}
+```
+
+Dataset item views receive `{ datasetItem, trace, context, actions }`. Use them for labeling expected output, metadata, review status, and scoring workflows.
+
+Available dataset actions:
+
+- `actions.updateDatasetItem({ expectedOutput?, metadata?, input?, sourceTraceId?, status? })`
+- `actions.createDatasetRunItem({ runName, datasetItemId?, traceId?, runDescription?, metadata? })`
+- `actions.createScore({ name, value, traceId?, comment?, metadata? })`
+
+When omitted, `datasetName`, `itemId`, and `traceId` default to the current Studio selection where Studio can infer them.

package/dashboard/.next/standalone/packages/kaizen/dashboard/.next/BUILD_ID

@@ -1 +1 @@
-YpQ-I4VL-aEdQrM5uN7_3
+SCF0o7YxElB9rzWaOohsA

package/dashboard/.next/standalone/packages/kaizen/dashboard/.next/build-manifest.json

@@ -4,8 +4,8 @@
   ],
   "devFiles": [],
   "lowPriorityFiles": [
-    "static/YpQ-I4VL-aEdQrM5uN7_3/_buildManifest.js",
-    "static/YpQ-I4VL-aEdQrM5uN7_3/_ssgManifest.js"
+    "static/SCF0o7YxElB9rzWaOohsA/_buildManifest.js",
+    "static/SCF0o7YxElB9rzWaOohsA/_ssgManifest.js"
   ],
   "rootMainFiles": [],
   "rootMainFilesTree": {},
@@ -15,9 +15,9 @@
       "static/chunks/framework-7089c270fe56b51f.js",
       "static/chunks/main-7ac7f96d288497aa.js",
       "static/chunks/431-43358ce3c29e5e1b.js",
-      "static/css/b18a6732b96168e1.css",
-      "static/chunks/673-ed4be46027ae7a37.js",
-      "static/chunks/pages/index-1d8b6719f49e4ae0.js"
+      "static/css/cd3873236eb77caa.css",
+      "static/chunks/253-85c76c34f33c9604.js",
+      "static/chunks/pages/index-d3306bb6f5d7d235.js"
     ],
     "/[system]": [
       "static/chunks/webpack-8c7966d82a2912f0.js",
@@ -30,45 +30,45 @@
       "static/chunks/framework-7089c270fe56b51f.js",
       "static/chunks/main-7ac7f96d288497aa.js",
       "static/chunks/431-43358ce3c29e5e1b.js",
-      "static/css/b18a6732b96168e1.css",
-      "static/chunks/673-ed4be46027ae7a37.js",
-      "static/chunks/pages/[system]/benchmarks-559dc9df52db3af4.js"
+      "static/css/cd3873236eb77caa.css",
+      "static/chunks/253-85c76c34f33c9604.js",
+      "static/chunks/pages/[system]/benchmarks-30a17b7659010b8c.js"
     ],
-    "/[system]/data": [
+    "/[system]/data/[[...path]]": [
       "static/chunks/webpack-8c7966d82a2912f0.js",
       "static/chunks/framework-7089c270fe56b51f.js",
       "static/chunks/main-7ac7f96d288497aa.js",
       "static/chunks/431-43358ce3c29e5e1b.js",
-      "static/css/b18a6732b96168e1.css",
-      "static/chunks/673-ed4be46027ae7a37.js",
-      "static/chunks/pages/[system]/data-644e4280b4c86fe0.js"
+      "static/css/cd3873236eb77caa.css",
+      "static/chunks/253-85c76c34f33c9604.js",
+      "static/chunks/pages/[system]/data/[[...path]]-e5f4083fe9ffe429.js"
     ],
     "/[system]/eval": [
       "static/chunks/webpack-8c7966d82a2912f0.js",
       "static/chunks/framework-7089c270fe56b51f.js",
       "static/chunks/main-7ac7f96d288497aa.js",
       "static/chunks/431-43358ce3c29e5e1b.js",
-      "static/css/b18a6732b96168e1.css",
-      "static/chunks/673-ed4be46027ae7a37.js",
-      "static/chunks/pages/[system]/eval-3c911ea8744631fd.js"
+      "static/css/cd3873236eb77caa.css",
+      "static/chunks/253-85c76c34f33c9604.js",
+      "static/chunks/pages/[system]/eval-160237a604b47416.js"
     ],
-    "/[system]/experiments": [
+    "/[system]/experiments/[[...path]]": [
       "static/chunks/webpack-8c7966d82a2912f0.js",
       "static/chunks/framework-7089c270fe56b51f.js",
       "static/chunks/main-7ac7f96d288497aa.js",
       "static/chunks/431-43358ce3c29e5e1b.js",
-      "static/css/b18a6732b96168e1.css",
-      "static/chunks/673-ed4be46027ae7a37.js",
-      "static/chunks/pages/[system]/experiments-42f31600c2bb47ad.js"
+      "static/css/cd3873236eb77caa.css",
+      "static/chunks/253-85c76c34f33c9604.js",
+      "static/chunks/pages/[system]/experiments/[[...path]]-91e47a4893093600.js"
     ],
     "/[system]/ideas": [
       "static/chunks/webpack-8c7966d82a2912f0.js",
       "static/chunks/framework-7089c270fe56b51f.js",
       "static/chunks/main-7ac7f96d288497aa.js",
       "static/chunks/431-43358ce3c29e5e1b.js",
-      "static/css/b18a6732b96168e1.css",
-      "static/chunks/673-ed4be46027ae7a37.js",
-      "static/chunks/pages/[system]/ideas-6829a271003150a9.js"
+      "static/css/cd3873236eb77caa.css",
+      "static/chunks/253-85c76c34f33c9604.js",
+      "static/chunks/pages/[system]/ideas-96e58e4624952e26.js"
     ],
     "/_app": [
       "static/chunks/webpack-8c7966d82a2912f0.js",

package/dashboard/.next/standalone/packages/kaizen/dashboard/.next/prerender-manifest.json

@@ -3,9 +3,9 @@
   "routes": {},
   "dynamicRoutes": {},
   "preview": {
-    "previewModeId": "02bf50b6d1c114ab64891ff63b9ae67b",
-    "previewModeSigningKey": "fe223618cfb9bb61306a9dea8261c44ddd7141789a6543970e870a61bf19da51",
-    "previewModeEncryptionKey": "a64d9df7a36d787a5996f6fb7171afb1fc8b7d35ad4962f94f482a0b687146c9"
+    "previewModeId": "0ba1834cfc7d7c8ea8708ad29269e503",
+    "previewModeSigningKey": "4cdb34c8c9deccef53be3f014ab0ccdd422a7b57e71290f1ce447fd0a2e2a138",
+    "previewModeEncryptionKey": "c48cc1326503410c45e8e0baf97d801d067e317fa759e378825440098271163f"
   },
   "notFoundRoutes": []
 }

package/dashboard/.next/standalone/packages/kaizen/dashboard/.next/routes-manifest.json

@@ -38,12 +38,13 @@
       "namedRegex": "^/(?<nxtPsystem>[^/]+?)/benchmarks(?:/)?$"
     },
     {
-      "page": "/[system]/data",
-      "regex": "^/([^/]+?)/data(?:/)?$",
+      "page": "/[system]/data/[[...path]]",
+      "regex": "^/([^/]+?)/data(?:/(.+?))?(?:/)?$",
       "routeKeys": {
-        "nxtPsystem": "nxtPsystem"
+        "nxtPsystem": "nxtPsystem",
+        "nxtPpath": "nxtPpath"
       },
-      "namedRegex": "^/(?<nxtPsystem>[^/]+?)/data(?:/)?$"
+      "namedRegex": "^/(?<nxtPsystem>[^/]+?)/data(?:/(?<nxtPpath>.+?))?(?:/)?$"
     },
     {
       "page": "/[system]/eval",
@@ -54,12 +55,13 @@
       "namedRegex": "^/(?<nxtPsystem>[^/]+?)/eval(?:/)?$"
     },
     {
-      "page": "/[system]/experiments",
-      "regex": "^/([^/]+?)/experiments(?:/)?$",
+      "page": "/[system]/experiments/[[...path]]",
+      "regex": "^/([^/]+?)/experiments(?:/(.+?))?(?:/)?$",
       "routeKeys": {
-        "nxtPsystem": "nxtPsystem"
+        "nxtPsystem": "nxtPsystem",
+        "nxtPpath": "nxtPpath"
       },
-      "namedRegex": "^/(?<nxtPsystem>[^/]+?)/experiments(?:/)?$"
+      "namedRegex": "^/(?<nxtPsystem>[^/]+?)/experiments(?:/(?<nxtPpath>.+?))?(?:/)?$"
     },
     {
       "page": "/[system]/ideas",
@@ -77,6 +79,12 @@
       "routeKeys": {},
       "namedRegex": "^/(?:/)?$"
     },
+    {
+      "page": "/api/langfuse-action",
+      "regex": "^/api/langfuse\\-action(?:/)?$",
+      "routeKeys": {},
+      "namedRegex": "^/api/langfuse\\-action(?:/)?$"
+    },
     {
       "page": "/api/langfuse-dataset",
       "regex": "^/api/langfuse\\-dataset(?:/)?$",
@@ -89,6 +97,12 @@
       "routeKeys": {},
       "namedRegex": "^/api/langfuse\\-dataset\\-item(?:/)?$"
     },
+    {
+      "page": "/api/langfuse-dataset-mutation",
+      "regex": "^/api/langfuse\\-dataset\\-mutation(?:/)?$",
+      "routeKeys": {},
+      "namedRegex": "^/api/langfuse\\-dataset\\-mutation(?:/)?$"
+    },
     {
       "page": "/api/langfuse-datasets",
       "regex": "^/api/langfuse\\-datasets(?:/)?$",
@@ -101,6 +115,12 @@
       "routeKeys": {},
       "namedRegex": "^/api/langfuse\\-trace(?:/)?$"
     },
+    {
+      "page": "/api/langfuse-traces",
+      "regex": "^/api/langfuse\\-traces(?:/)?$",
+      "routeKeys": {},
+      "namedRegex": "^/api/langfuse\\-traces(?:/)?$"
+    },
     {
       "page": "/api/linear-ideas",
       "regex": "^/api/linear\\-ideas(?:/)?$",
@@ -150,8 +170,8 @@
       "routeKeys": {
         "nxtPsystem": "nxtPsystem"
       },
-      "dataRouteRegex": "^/_next/data/YpQ\\-I4VL\\-aEdQrM5uN7_3/([^/]+?)\\.json$",
-      "namedDataRouteRegex": "^/_next/data/YpQ\\-I4VL\\-aEdQrM5uN7_3/(?<nxtPsystem>[^/]+?)\\.json$"
+      "dataRouteRegex": "^/_next/data/SCF0o7YxElB9rzWaOohsA/([^/]+?)\\.json$",
+      "namedDataRouteRegex": "^/_next/data/SCF0o7YxElB9rzWaOohsA/(?<nxtPsystem>[^/]+?)\\.json$"
     }
   ],
   "rsc": {