CodeMeridian.Indexer 1.0.38

CodeMeridian Indexer

CodeMeridian.Indexer is the unified CLI for indexing code into CodeMeridian from C#, TypeScript/TSX, documentation, and configuration sources.

Install

dotnet tool install -g CodeMeridian.Indexer

Use

The indexer sends data to a running CodeMeridian backend. Run codemeridian serve in a dedicated runtime folder for the shared MCP server and Neo4j stack, then run codemeridian init . in each indexed project or codemeridian init --global for user-wide defaults.

To create local runtime files and start Neo4j plus the MCP server with Docker Compose:

codemeridian serve --no-start
codemeridian serve

codemeridian init .
codemeridian index .
codemeridian index C:\Projects\MyApp --project MyApp --clear
codemeridian index .
codemeridian index . --skip-keywords
codemeridian keywords --project MyApp
codemeridian keywords rebuild --project MyApp
codemeridian keywords index --project MyApp
codemeridian keywords classify --project MyApp
codemeridian keywords rebuild --project MyApp
codemeridian keywords status --job-id 11111111-2222-3333-4444-555555555555
codemeridian config rebuild --project MyApp
codemeridian index . --skip-csharp --skip-docs --skip-diagnostics
codemeridian index . --skip-config
codemeridian index . --watch

codemeridian doctor --project CodeMeridian
codemeridian report --project CodeMeridian
codemeridian report pr-context --project CodeMeridian --base origin/main --head HEAD --format markdown --output artifacts/pr-context.md
codemeridian check-drift --project CodeMeridian --fail-on high
codemeridian evaluate-session . --project MyApp --session .meridian/sessions/session.jsonl

What It Does

• Detects C# projects, TypeScript/TSX roots, and documentation files.
• Detects repo-local configuration files such as appsettings*.json, meridian*.json, .env, and Docker Compose YAML.
• Indexes code into Neo4j through CodeMeridian.
• Indexes canonical configuration keys and links direct and typed C# configuration usage into the same graph.
• Skips unchanged files after the first successful run using .meridian/cache.
• Can run compiler, TypeScript, and lint diagnostics unless you skip them.
• Rebuilds and classifies the backend keyword graph after indexing unless --skip-keywords is set.
• Queues incremental keyword refresh automatically as nodes and documents are ingested by the MCP server.
• Can rebuild the keyword graph on demand without indexing through codemeridian keywords. Rebuild also runs keyword classification after the rebuild finishes.
• Can classify already-built keywords on demand through codemeridian keywords classify.
• Can rebuild the configuration graph on demand without rerunning language indexers through codemeridian config rebuild.
• Can read configurationFiles from meridian.json to control which config file names or wildcard patterns are indexed.
• Repo-controlled build and lint diagnostics are opt-in via --allow-repo-scripts.
• Can query the backend for a doctor status report without talking to Neo4j directly from the client.
• Can print an architecture weather report with graph counts, risks, bridge nodes, coverage gaps, and freshness.
• Can generate CI-friendly PR context reports as Markdown or JSON without requiring an interactive assistant.
• Can verify graph drift with codemeridian check-drift or codemeridian index --verify.
• Can evaluate whether CodeMeridian helped an implementation session by comparing provider-neutral session evidence with git changes, then write .meridian/precision-feedback.json for future ranking feedback.
• Can create local runtime files and start the backend stack with codemeridian serve.
• Supports dry runs and capability listing for environment checks.
• Can generate a local meridian.json and MCP client config with an auto-detected project name.
• Can refresh an existing meridian.json in place by rerunning codemeridian init ., merging missing defaults without overwriting local settings.
• Can seed .meridian/architecture.json and copy bundled templates from the package architectures/ folder into .meridian/architectures/ so architecture checks are project-owned and editable.
• Can seed .meridian/keyword-classification.json from the packaged root keyword-classification.sample.json so keyword classification rules live in the repo instead of being hardcoded in the backend.
• Can seed .meridian/database-tracing.json from the packaged root database-tracing.sample.json so database-recognition presets stay repo-owned instead of being buried in backend code.
• Can copy bundled agent guidance into meridian-agent-capabilities/ so repo-local agent instructions travel with the setup without writing into a user-owned docs/ tree, including Codex-specific helper scripts under codex-scripts/.
• Can also read CodeMeridian_Project from .env when you want a fixed project name without --project.

Package Contents

• LICENSE
• This readme
• The bundled TypeScript indexer assets

Notes

• Use --project <name> when you want a stable project context.
• Use CodeMeridian_Project in .env when you want the same project context applied automatically.
• Use codemeridian doctor --project <name> to ask the backend for graph health, drift, and counts.
• Use codemeridian report --project <name> for a compact architecture weather report.
• Use codemeridian trace-endpoint "POST /api/orders" --project <name> when you want a graph-only route trace through indexed database and messaging paths. Database tracing is driven by .meridian/database-tracing.json and currently ships starter presets for EF Core, Dapper, raw SQL, Prisma, Knex, and Neo4j Cypher. trace_endpoint is an alias if you prefer the MCP-style name.
• Use codemeridian report pr-context --base <git-ref> --head <git-ref> when you want a deterministic PR review summary with changed graph nodes, impact hints, missing-test warnings, hotspot/churn warnings, and related docs. Add --format json for automation and --output <path> for CI artifacts.
• Use codemeridian check-drift --project <name> --fail-on high for a drift gate that exits non-zero in CI when the graph is too stale.
• Use codemeridian evaluate-session --project <name> to read the newest .meridian/sessions/*.jsonl evidence file, report whether CodeMeridian suggestions matched changed files and tests, and refresh .meridian/precision-feedback.json. Pass --session <file-or-directory> to choose evidence explicitly and --base <git-ref> to change the diff base. See Evaluating Session Usefulness for the step-by-step workflow.
• Use codemeridian index --verify --project <name> when you want the same drift gate as part of an indexer workflow.
• Use --skip-diagnostics if you only want structural indexing.
• Use --skip-config when you want to skip configuration-file parsing and config-usage graph edges for a run.
• Use --skip-keywords when you do not want the index run to finish by rebuilding and classifying derived keywords.
• Use codemeridian config rebuild --project <name> when you want a clean rebuild of configuration-file nodes, canonical config keys, and C# config usage links.
• Use .meridian/database-tracing.json to tune how C# indexing recognizes EF Core, Dapper, and raw SQL database access. The default presets emit DatabaseOperation and DatabaseTable graph concepts without hardcoding one data-access style into the backend.
• Use codemeridian keywords --project <name> as the short form when you want a rebuild plus classification. It submits a background job and returns a job id immediately.
• Use codemeridian keywords rebuild --project <name> when you want an explicit maintenance command name. It also submits a background job and runs classification after rebuild.
• Use codemeridian keywords index --project <name> if you prefer index terminology; it is an alias of rebuild.
• Use codemeridian keywords classify --project <name> when you only want to classify existing derived keywords without rebuilding relationships. It also submits a background job.
• Use codemeridian keywords rebuild --project <name> --wait or codemeridian keywords classify --project <name> --wait when you want to start the job in the background but keep the CLI attached until it completes.
• Use codemeridian keywords status --job-id <id> when you want to verify whether a background keyword job finished, failed, or expired.
• Use --allow-repo-scripts only on trusted repos when you want dotnet build and repo lint commands to run.
• Use --no-incremental or --force-full to scan all files without clearing the project.
• Use codemeridian init . to create or refresh meridian.json for a project and then step through prompts for .vscode, .continue, .codex, and meridian-agent-capabilities. When meridian.json already exists, init merges missing defaults, bumps the config version, and writes meridian.json.bak before replacing the file. The generated meridian.json enables allowRepoScripts by default for trusted repos.
• Use codemeridian init --global --url http://localhost:5100 to create a user-level fallback config when you want the CLI to work across many repos without project-local config. Global init also seeds .meridian/architecture.json, .meridian/keyword-classification.json, .meridian/database-tracing.json, .meridian/architectures/, and meridian-agent-capabilities/ under the global config root.
• Use codemeridian serve in a separate runtime folder to create .env and docker-compose.codemeridian.yml, then run docker compose pull followed by docker compose up -d.
• Use codemeridian serve --no-start when you only want to write or merge those runtime files.
• Use codemeridian init . when you want to create or merge client MCP config such as .vscode/mcp.json, .continue/mcpServers/code-meridian.yaml, or .codex/config.toml.
• On Windows, keep CodeMeridian_Auth_ApiKey in the runtime .env for the server and also in User or System environment variables for MCP clients such as VS Code or Codex; project .env files are not automatically imported into those client processes.
• The global tool does not include Neo4j. It starts Neo4j and the MCP server through Docker using the published MCP server image.
• The repo-level README covers the full CodeMeridian product and architecture.