M1 5t internal pilot ready

Due Mar 31, 2026

Current68 / 97 (70%)

Exit criteria (0/5)

• ☐ M1 pilot feature bundle is complete and usable across key surfaces: extension + profiler, MCP/Copilot authoring flow, Suite auth/permissions/git/authoring, and AI_CONTEXT beta contract.
• ☐ 5T Analytics raw-source connectivity is live and validated end-to-end in the hosted Fivetran pilot environment.
• ☐ Internal analysts can run weekly real workflows without one-off engineering intervention.
• ☐ Pilot operations are in place: instrumentation, weekly triage cadence, runbooks, and owner-based closure tracking.
• ☐ Pilot readiness decision is explicit: go/no-go recorded with risks, open gaps, and next actions.

KPI targets (0/5)

• ☐ Analyst activation: at least 5 pilot analysts onboarded, with at least 3 weekly active analysts for 2 consecutive weeks.
• ☐ Workflow success rate: at least 80% of tracked pilot workflows complete end-to-end without engineering intervention.
• ☐ Time to value: median time from opening product to first working dashboard is 30 minutes or less.
• ☐ Reliability: no P0 incidents and no unresolved P1 incident older than 5 business days during the pilot window.
• ☐ Pilot sentiment: at least 80% of pilot analysts give a clear thumbs-up to proceed to broader internal rollout.

Tasks by Workstream

dft core (Sr Engineer Architect)

Internal analysts can execute at least one weekly real workflow that depends on the YAML contract, compiler/normalizer, execution adapters, and release/versioning in the 5T Analytics environment, without bespoke engineering intervention for every run. Instrumentation and feedback capture are in place so failures, friction points, and adoption gaps are visible and triaged with owners.

• ☑ Remove init_sql and migrate setup patterns to query composition — Deprecate and remove init_sql/init_sql_file, then migrate setup use cases to reusable query imports (for example `_…
• ☑ Add chart renderer registry and mechanical Vega renderer — Introduce renderer-family dispatch, rebuild standard Vega-Lite translation around resolved chart semantics, and make la…
• ☑ Build chart intent and enrichment pipeline — Introduce ChartIntent, EnrichmentPatch, ResolvedChart, RenderArtifact, and the authored/enriched/config precedence flow…
• ☑ Document faces/ directory convention and dbt project setup — Fix all docs to use faces/ as the canonical dashboard directory. Add Adding Dataface to an Existing dbt Project guide t…
• ☑ Move chart output conversion into render converters — Relocate chart export conversion out of chart semantics and route SVG, PNG, and PDF conversion through render/converter…
• ☑ Rebuild geo renderer on resolved chart pipeline — Adapt geo rendering to the new intent/enrichment/resolution pipeline with shared source, join, projection, and tooltip…
• ☑ Refactor chart renderer boundaries and dispatch — The chart rendering layer still mixes orchestration, Dataface JSON serialization, SVG-native export conversion, and Veg…
• ☑ refactor: Move database/source detection from playground to core — Move database/source detection into dft core so playground and hosted surfaces share one canonical detection path.
• ☑ Deduplicate geo chart rendering and config ownership — The geo rendering code still duplicates point-map and layered-point-map behavior across geo source resolution, projecti…
• ☑ Extract shared Vega-Lite spec builder helpers — Chart-specific Vega-Lite generators still repeat the same spec assembly patterns for title handling, dimensions, toolti…
• ☑ Tighten chart enrichment and type inference boundaries — Chart enrichment and type inference still cross module boundaries in ways that blur responsibilities between render-tim…

cloud suite (UI Design and Frontend Dev)

Internal analysts can execute at least one weekly real workflow that depends on hosted user experience for onboarding, sharing, collaboration, and account/project flows in the 5T Analytics environment, without bespoke engineering intervention for every run. Instrumentation and feedback capture are in place so failures, friction points, and adoption gaps are visible and triaged with owners.

• ☑ Embed AI agent and MCP actions in dashboard/chart builder pages — Integrate AI agent and MCP actions directly into Suite dashboard and chart creation pages for pilot analyst workflows.
• ☐ Suite chart creation and settings editing UX pass — Improve rough chart creation and chart settings editing UX in Suite to a pilot-usable baseline for internal analysts.
• ◔ Suite Google login for pilot — Productionize Google login in Cloud Suite as the primary account authentication path using the existing django-allauth…
• ◔ Split shared cloud chat UI into stream render and format modules — Refactor apps/cloud/static/js/chat/chat.js into smaller modules that separate SSE transport, message/tool rendering, an…

inspect profiler (Sr Engineer Architect)

Internal analysts can execute at least one weekly real workflow that depends on warehouse profiling, semantic inference, and analyst-facing data context surfaces in the 5T Analytics environment, without bespoke engineering intervention for every run. Instrumentation and feedback capture are in place so failures, friction points, and adoption gaps are visible and triaged with owners.

• ☑ Enable profiler drill-in/out links across table, schema, and column dashboards — Add working navigation links so analysts can move from table profiles into schema/column dashboards and back out withou…
• ◔ Profiler payload and UX contract ready for extension consumption — Stabilize inspect/profiler output contract and UX assumptions so extension embedding is reliable for M1 pilot.
• ☑ Add histogram bins and date distributions to profiler — Add profiler histogram bins and date distributions so analysts can understand value spread and temporal density at a gl…
• ☑ Add spark bar chart type for profiler column cards — Add spark bar chart support for profiler column cards to improve compact distribution and completeness scanning.
• ☐ IDE inspector: use cached inspect.json before querying database — All profiler surfaces should read from inspect.json as single source of truth. Never auto-profile — prompt user on cach…
• ◔ Inspector cleanup wave 1 architectural decomposition and contract hardening Inspector cleanup and open-source hardening — Plan and execute a deeper inspector cleanup pass that decomposes oversized inspector modules, tightens internal APIs, i…
• ☑ Inspector: fetch and display database column comments — Fetch and display database column comments in inspector so semantic context from warehouses is visible during analysis.
• ☑ Eliminate all custom HTML - dataface YAML everywhere — Replace all hand-crafted HTML across the extension and server with dataface YAML rendered through the normal compile/re…
• ☑ Refactor TableInspector inspection pipeline for maintainability — Reduce the complexity of TableInspector._inspect_table_inner by extracting private helpers while preserving profiler be…

mcp analyst agent (Data AI Engineer Architect)

Internal analysts can execute at least one weekly real workflow that depends on AI agent tool interfaces, execution workflows, and eval-driven behavior tuning in the 5T Analytics environment, without bespoke engineering intervention for every run. Instrumentation and feedback capture are in place so failures, friction points, and adoption gaps are visible and triaged with owners.

• ☑ Extract shared chat.js and chat_stream SSE endpoint — Extract the shared chat component chat.js and chat_stream SSE endpoint as a standalone M1 task. This resolves the depen…
• ☑ MCP tooling contract for extension + Copilot dashboard/query generation — Define and harden MCP tool inputs/outputs so extension and Copilot can reliably generate dashboards and queries in pilo…
• ☑ Unify Cloud AI Tool Dispatch to Use Canonical MCP Tools AI Agent Surfaces — Replace the bespoke _execute_tool_sync() in apps/cloud/apps/ai/views.py (which only supports 4 tools: validate_yaml, te…
• ☑ Wire Playground AI to use MCP tools instead of bespoke tool set — The Playground app currently maintains its own bespoke AI tools - validate_yaml, test_yaml_execution, execute_query_res…
• ☑ Add JSON render output format — Add format=json to the render pipeline that walks the layout tree, executes queries, resolves charts, and returns the r…
• ☑ Refactor Cloud AI chat stream into scoped execution services — Refactor apps/cloud/apps/ai/views.py chat_stream into smaller scope-resolution, tool-execution, and SSE-streaming units…
• ☑ Replace AI tool dispatch switch with registry-backed handlers — Refactor dataface/ai/tools.py so canonical tool schemas and handlers are registered in one place instead of maintained…
• ☑ Save dashboard MCP tool - persist agent work to project — Add a save_dashboard MCP tool that writes agent-generated YAML to the project file system. Currently all tools are stat…
• ☑ Scope playground MCP surface to playground sources — Refactor the shared AI/MCP surface to accept an injected context for adapter registry, dashboard directory, base dir, a…
• ☐ Wire Dataface to internal analytics repo and BigQuery source — Set up the Dataface-side access path to the internal analytics warehouse and sibling analytics dbt repo. Use /Users/dav…
• ☐ Add resolved YAML render output format — Add a format=yaml output that produces a resolved dataface YAML -- auto chart types filled in, auto-detected fields exp…
• ☑ Type terminal agent event protocol and provider stream adapters — Refactor the terminal agent loop introduced in dataface/ai/agent.py and dataface/ai/llm.py to use explicit typed event…

ft dash packs (Data Analysis Evangelist and AI Training)

Internal analysts can execute at least one weekly real workflow that depends on connector-specific dashboard packs and KPI narratives for Fivetran sources in the 5T Analytics environment, without bespoke engineering intervention for every run. Instrumentation and feedback capture are in place so failures, friction points, and adoption gaps are visible and triaged with owners.

• ☑ Populate Faketran application database models for fake companies — Audit vendored Faketran fake-company sources for internal application database coverage, then populate and validate the…
• ☐ Transform mockusign_dbt into realistic dbt project with staging/marts — Upgrade mockusign_dbt into a realistic staged/marts dbt project suitable for internal analyst pilots and demos.
• ☑ Vendor faketran as a monorepo lib and replace mockusign/gruber datasets — Pull faketran into this monorepo as a repo-owned library/module and replace the current mockusign/gruber example data…

ide extension (Head of Engineering)

Internal analysts can execute at least one weekly real workflow that depends on analyst authoring workflow in VS Code/Cursor with preview, diagnostics, and assist in the 5T Analytics environment, without bespoke engineering intervention for every run. Instrumentation and feedback capture are in place so failures, friction points, and adoption gaps are visible and triaged with owners.

• ☐ IDE extension usability hardening for internal analyst pilot — Harden extension UX so YAML editing, validation, preview, and error recovery are smooth enough for non-engineer interna…
• ☐ Profiler surface in IDE extension is pilot-ready — Deliver profiler and inspector surfaces in VS Code/Cursor extension with stable rendering, interactions, and fallback b…
• ☐ Enable internal analyst IDE flow — Ship a stable VS Code/Cursor flow for YAML editing, validation, and preview in pilot usage.
• ☑ Open in browser regression coverage and shell-safe serve launch — Add end-to-end openInBrowser regression tests and harden VS Code terminal command construction for project-dir paths.
• ◔ Split VS Code inspector runtime from webview controller — Decompose apps/ide/vscode-extension/src/inspector/inspector-panel.ts by extracting server/runtime orchestration, naviga…

graph library (Data Visualization Designer and Engineer)

Internal analysts can execute at least one weekly real workflow that depends on visual language, chart defaults, interaction behavior, and differentiated styling in the 5T Analytics environment, without bespoke engineering intervention for every run. Instrumentation and feedback capture are in place so failures, friction points, and adoption gaps are visible and triaged with owners.

• ◔ M1 consistent category color mapping across charts — Investigate and design a top-level category color mapping system so repeated semantic values such as Accessories, Elect…

context catalog nimble (Data AI Engineer Architect)

Internal analysts can execute at least one weekly real workflow that depends on context schema/catalog contracts and Nimble enrichment flows across product surfaces in the 5T Analytics environment, without bespoke engineering intervention for every run. Instrumentation and feedback capture are in place so failures, friction points, and adoption gaps are visible and triaged with owners.

• ☑ AI_CONTEXT beta health and readiness scorecard — Define and track AI_CONTEXT beta health metrics so M1 go/no-go is based on coverage, quality, and analyst usability sig…
• ☑ AI_CONTEXT layer 6 relationship mapping for pilot datasets Layer 6 Relationship Mapping — Implement cross-table relationship mapping in AI_CONTEXT so join graph context is available to agents during M1 workflo…
• ☑ AI_CONTEXT metadata contract v1 for pilot MCP Catalog and Agent Tools — Solidify AI_CONTEXT data format into a versioned contract with clear field semantics and compatibility rules for beta u…
• ◔ Description priority merge in MCP context output Description Enrichment Pipeline — Implement deterministic description-source merging in MCP context output so AI tools receive stable best-available sema…
• ☑ Ingest dbt schema.yml descriptions into AI_CONTEXT Description Enrichment Pipeline — Merge dbt model and column descriptions into AI_CONTEXT so human-authored semantics are available during pilot analysis.
• ☑ AI_CONTEXT grain and fanout risk signals (beta subset) Grain Inference and Fanout Risk — Ship grain candidate, join multiplicity, and fanout risk metadata in AI_CONTEXT to reduce unsafe aggregate query genera…
• ◔ dft inspect native CSV support via ephemeral DuckDB — dft inspect cannot profile CSV sources today because the inspector only supports SQL databases. It should handle CSVs n…
• ☑ dft inspect: build complete self-contained catalog in target/inspect.json — dft inspect should be the single command that builds a complete, self-contained catalog artifact in target/inspect.json…
• ◔ Incremental dft inspect with lineage-aware change detection — dft inspect should skip re-profiling tables whose source data and upstream lineage have not changed since the last insp…
• ◔ Move playground examples to DuckDB and ship pre-built inspect.json — Playground examples currently use raw CSV files via CsvAdapter with Python stdlib csv.DictReader - no SQL, no joins, no…
• ☑ search_dashboards MCP tool for pilot context workflows MCP Catalog and Agent Tools — Add search_dashboards MCP tool so pilots can discover relevant existing dashboards and reuse validated query patterns.

dashboard factory (Data Analysis Evangelist and AI Training)

Internal analysts can execute at least one weekly real workflow that depends on repeatable process for producing, reviewing, and publishing quickstarts/examples in the 5T Analytics environment, without bespoke engineering intervention for every run. Instrumentation and feedback capture are in place so failures, friction points, and adoption gaps are visible and triaged with owners.

• ☑ Add dashboard review-and-revise workflow — Define and pilot a second-pass dashboard review workflow that inspects rendered dashboards with real data, captures rev…
• ☐ Add dft init for dbt-native repo bootstrap — Implement a first-class dft project bootstrap command for existing repos, especially dbt repos that will store dashboar…
• ☑ Add master plans daily activity page — Track completed tasks by day with owners, completers, and linked PRs, including merged PRs not tied to tasks.
• ◔ Build master_plans DuckDB SQL metrics pipeline for milestone dashboards — Export planning data to Parquet, query via DuckDB, and drive milestone header visualizations from SQL so progress views…
• ☐ Create analytics repo Dataface branch and bootstrap workflow — Set up the internal analytics repo as a first-class Dataface example-customer repo for analyst work. Create and documen…
• ◔ Own Vega-Lite schema snapshot and chart defaults — Vendor the Vega-Lite schema as a tracked compile-time artifact, add a dedicated chart defaults YAML for Dataface house…
• ☐ Add render command for precomputed dashboard data artifacts — Add a CLI path for rendering from a precomputed JSON artifact so the final render stage can be rerun without recompilin…

infra tooling (Sr Engineer Architect)

Internal analysts and engineers can execute at least one weekly real workflow that depends on developer tooling, cbox runtime reliability, and deployment execution safety in the 5T Analytics environment, without bespoke engineering intervention for every run. Instrumentation and feedback capture are in place so failures, friction points, and adoption gaps are visible and triaged with owners.

• ☑ Harden cbox sandbox bootstrap with PATH, pre-commit, and git auth health checks — Eliminate recurring sandbox runtime failures by standardizing environment bootstrap and adding startup health checks fo…
• ◔ Host Dataface on Fivetran GCP — Stand up and harden the canonical GCP runtime path for internal pilot usage, including deploy auth hardening and intern…
• ☑ Add full-name output mode for cbox list — Prevent manager confusion from truncated session names by supporting a full-width/raw list mode.
• ☑ Add per-worktree local port bundles for dispatch QA — Allocate unique local serve ports for each dispatched worktree, write a worktree-local ports file similar to .cbox-port…
• ☑ Align cbox bootstrap/health docs with actual commands — Fix canonical and wrapper docs where bootstrap and git auth health-check commands diverge from implementation.
• ☑ cbox manager default parent-branch policy — Make manager-launched sandboxes default to the manager's active branch (not main) unless explicitly overridden with --p…
• ☑ CBox manager interactive stall detection and recovery — Detect and recover manager flows stuck at interactive prompts (e.g. /pr-lite menus, blocker prompts) with deterministic…
• ☑ CBox Process: diagnose and hard-fail silent cbox review failures in sandboxes — Observed in M1-INFRA-027 forensic run: sandbox had CBOX_CONTAINER=1 so cbox review should use _run_review_in_tmux, but…
• ☑ CBox Process: hard-block PR when cbox review runtime is missing — Observed during M1-INFRA-027: sandbox /pr flow offered 'skip review, open PR' when Docker/Podman missing, leading to PR…
• ☑ CBox review prompt context isolation on sandbox restart — Observed restart path where sandbox opened with stale review prompt context (.cbox/.review-prompt.md flow). Ensure sand…
• ☑ CBox sandbox bootstrap health parity for python and pre-commit — Repeatedly observed on fresh sandbox start: bootstrap health checks fail for python and pre-commit immediately after se…
• ☑ CBox sandbox git metadata path isolation — Fix sandbox git commands failing because worktree metadata points to host paths (e.g., '/Users/.../.git/packed-refs').…
• ☑ CBox sandbox session liveness drop detection and recovery — Observed manager incident: sandbox session disappeared ('No session found') while worktree/branch remained intact. Add…
• ☑ cbox sandbox sessions can exit unexpectedly during long task handoff — Track issue 423 in master_plans after retiring GitHub Issues as the active backlog.
• ☑ CBox sandbox startup-timeout diagnostics — Surface actionable diagnostics when wait_for_prompt times out during sandbox or review startup, replacing opaque "Timeo…
• ☑ cbox send false-positive delivery when sandbox TUI ignores input — Investigate and fix cases where cbox send reports success after tmux send-keys, but the target sandbox Claude TUI does…
• ☑ CBox session registry stale after sandbox kill — Observed during manager cleanup: 'cbox new --list' continued showing a killed sandbox while 'cbox list' showed no sessi…
• ☑ CBox setup-worktree ROOT_WORKTREE_PATH fallback — Harden worktree setup so cp from root .env succeeds when ROOT_WORKTREE_PATH is unset by deriving the root path from git…
• ☑ Configurable review timeouts and stall detection — Add configurable review timeout (CLI flag + env var) with 20m default, and stall detection that distinguishes slow-but-…
• ☑ Decouple task workflow from cbox CLI — add /cbox-task command surface — Keep core cbox generic and make task workflow optional via a composable skill/command layer.
• ☑ Improve cbox recovery from hung in-session tool calls — Ensure manager interrupt/send can reliably recover sandboxes stuck in long-running shell tool calls.
• ☑ Make cbox entrypoint bootstrap timeout configurable — Replace hard-coded bootstrap timeout values with environment-configurable settings and sane defaults.
• ☐ Make qa-explorer use local browser subagent without cbox fallback — Make qa-explorer run through the local subagent/browser path instead of any cbox fallback, ensure the browser automatio…
• ☑ Master Plans CLI ergonomics and command wrappers — Make master_plans task tooling easier to run than raw python invocation by adding a user-friendly command entrypoint, c…
• ☑ Prevent cbox sandboxes from mutating host git common-dir — Sandbox containers currently mount the host repo common .git directory writable at /workspace/.repo-git. Diagnose and f…
• ☑ Reduce cbox sandbox startup latency by parallelizing health checks — Run post-boot sandbox health checks concurrently instead of sequentially to reduce worst-case startup delay.
• ☐ Scope just server bindings by execution context — Make localhost the default host binding for local recipes and keep 0.0.0.0 where container access requires it.
• ☑ Simplify PR checklist enforcement and reduce brittle PR body sync — Reduce false-negative PR checklist failures by removing redundant gates, consolidating label definitions, and making en…
• ☑ Add cbox test command for running visual tests locally in Linux container — Add a repeatable cbox command that runs visual tests locally in the Linux container used by CI.
• ◔ Consolidate local dispatch and review scripts behind shared implementation — Refactor the new local worktree dispatch and review tooling so scripts/dispatch, scripts/dispatch-kill, scripts/review,…
• ☑ Improve merge flow guidance for worktree-bound local branches — Document and handle expected local branch deletion warnings after gh pr merge --delete-branch when branch is checked…
• ☑ Smarter cbox cleanup: detect squash-merged PRs, ignore sandbox artifacts, scan all worktrees — Make cbox cleanup catch squash-merged branches, ignore .claude-sessions-sandbox/ as dirty state, check commits-ahead, a…

integrations platform (Head of Engineering)

Internal analysts can execute at least one weekly real workflow that depends on deployment, billing, 5T connectivity, and operational reliability/launch integration in the 5T Analytics environment, without bespoke engineering intervention for every run. Instrumentation and feedback capture are in place so failures, friction points, and adoption gaps are visible and triaged with owners.

• ☐ Enable 5T Analytics connectivity for analyst pilot — Connect 5T Analytics data path, secrets, and environment configuration for pilot analyst workflows.
• ◔ Operationalize Fivetran GCP deploy path — Finalize deployment pipeline/runbooks on Fivetran Google Cloud infrastructure for pilot reliability.