Skip to main content

Roadmap: v2 vision

GhostMap Core (everything described in the rest of this documentation) is the V1 scope under the GhostMap Free Non-Commercial License (source-available, non-commercial use). What follows is the vision for future Enterprise capabilities, aimed at teams where this information needs to flow between the technical, operational, and management layers. Commercial use would require written authorization or a future licensing flow.

Nothing in this section is shipped: Everything described here (Ghost Index v2, Ghost Watcher, Ghost Threads, Ghost Graph, dashboards, AI explanations, Jira / Slack integrations, permissions / audit log) is future direction, not functionality available in V1 (0.5.x). What already works today is described in the rest of the documentation.

Any future integration with server services will require its own disclosure in GhostMap's Privacy Policy and explicit consent before any data leaves your machine. For a current copy, write to getghostmap@proton.me. Nothing in this roadmap is a commitment on dates or final scope.

V2: workspace-wide indexing engine (planned)​

State today (V1, 0.5.x): the Ghost Tree is computed per active file. Every open file persists its snapshot to .ghostmap/ghostmap.json (see Local State). That is a per-document cache, not a workspace index. There is no filesystem watcher and no background fingerprint validator (the earlier attempt was removed because it froze the Extension Host).

V2 direction (not shipped): a persistent workspace-level index (the Ghost Index v2) built with JSON shards per top-level folder, kept up to date incrementally by a Ghost Watcher on the filesystem and validated out of band by a Ghost Validator in a worker thread (mtime prefilter + streaming SHA-256). The active-file refresh path (V1 responsiveness) is preserved: the V2 engine layers in below, it does not replace it.

Benefit if shipped: opening any file in the workspace in milliseconds from the index, cross-file navigation, and a base for richer integrations. Until it ships, this is roadmap direction.

Potential index content: files, symbols, anchors, metadata, resolved hierarchy, diagnostics, and, in the future, file-to-file relations / dependencies (the Ghost Graph, also known as the Ghost Context Graph).

Language expansion (separate workstream, planned)​

Independent of the V2 engine, a workstream exists to add roughly 20 more languages (Kotlin, Swift, Haskell, OCaml, Clojure, Lua, R, Bash, the SQL family, and others) on top of today's 19. Not shipped. The gate before any language is marked supported is:

  • reproducible packaging of the Tree-sitter / WASM grammars,
  • query validity against the exact grammar version,
  • a "load + sample query" smoke per grammar,
  • nesting / icons / anchors fixtures in test/matrix/,
  • and, where required, upstream PRs to the grammar.

See Disclaimer: Language pack expansion and the "Language expansion" section in Requirements.

Gate before future integrations: Before publishing any integration with Jira, Slack, AI, dashboards, or server services, GhostMap has to close these points as product requirements, not as optional details:

  • Permissions and roles: which actions a person can initiate, which an agent can suggest, and which require explicit approval.
  • Data model: which fields are stored locally, which travel to external services, how they are versioned, and how they are deleted.
  • Consent and privacy: an updated notice, in-editor consent before transmitting data, and a clean split between V1 local-only and future connected capabilities.
  • Recovery tests: scenarios for revoked permissions, expired tokens, downed integrations, write conflicts, and safe rollback if an automation fails.

Until those gates exist and are verified, the connected capabilities stay vision, not shipped product.

Pillar 1: Resume work without friction​

A dev finishes the day. The next day, they open VS Code and GhostMap.

  • General dashboard showing the state of every file and the Ghosts marked as urgent.
  • Urgency derived from Jira: GhostMap connects to Jira, reads sprint timing for the active sprint and dependent sprints, and (with AI help) identifies relations and dependencies between sprints to flag urgent Ghosts (by their own due date or because they block another sprint).
  • Local chat in VS Code that summarizes relevant conversations and context, and proposes what to do next based on urgency and scope.
  • One-click resume: the dev jumps straight into the first urgent task, or continues exactly where they left off the day before.
  • Ghost Threads (per-block discussions): any function, class, or block can have its own discussion, with a technical owner, QA, etc., with configurable permissions, including comments sent from Slack into a specific discussion.
  • Tangible progress in Jira: when analyzing a Ghost or a large Range Anchor, GhostMap can propose a decomposition into Jira subtasks. Each subtask links back to its own Ghost in code, and as they are marked done, the parent task's progress updates proportionally and visibly.
  • Automated task closure: when a Ghost's status flips to done (or equivalent), GhostMap can generate the matching commit, update Jira, notify QA/reviewer, and reflect the new state in real time for the whole team.
  • Permissions and traceability: because this layer can generate commits, modify tickets, and create subtasks automatically, Enterprise includes role-based permissions and a record of which action was taken by a person and which by an AI agent.

Pillar 2: Understand someone else's code​

A new dev opens a file and does not know what is actually there.

  • AI-generated explanations for symbols without prior documentation.
  • Ghost Graph: graphs showing relations between files, symbols, and dependencies (the Ghost Context Graph from the index).
  • Lists with context and urgency: while viewing what functions / classes a file holds, the urgent items in that same file are also flagged.
  • Decision history: direct access to past discussions about why a decision was made in that block of code. Information that today only lives in the head of whoever wrote it.
  • Non-mandatory structure suggestions: GhostMap can suggest organizing the code (for example with Range Anchors) without forcing it.
  • In-editor notifications: status changes or new relevant discussions arrive in VS Code, one click away from the exact spot.

Ghost Comments and retroactive documentation​

  • Ghost Comments (invisible ghosts): in v2, Ghost information stops living as literal text in the file. VS Code shows it visually with decorations and CodeLens, without taking up lines in the source. The code stays clean and the information lives in the Ghost Index v2.
  • Conversion of existing comments: TODO/FIXME style comments already in the code can be converted automatically into Ghost Metadata, reusing the same ownership resolution that exists today.
  • Automatic documentation of legacy code: AI and graph analysis over code with no Ghosts at all, which proposes automatic Range Anchors and generates descriptions. Useful for "bootstrapping" the index on projects that never used GhostMap.

Next step​

For the current V1 state and known limits, continue with Project status.

Last updated on