first cut at re-naming the fork

.gitignore

@@ -1,9 +1,8 @@
 # Generated artifacts
-codeindex.json
+blastradius.json
 symbolindex.json
 repo_graph.json
-symbolindex.json
-.codeindex/
+.blastradius/
 benchmark/results.json
 
 # Claude Code

AGENTS.md

@@ -0,0 +1,36 @@
+# AGENTS.md
+
+## Developer Commands & Verification
+
+- **Run tests**: `uv run pytest` (set `$env:PYTHONPATH="."` if running outside an installed venv).
+- **Run single test**: `uv run pytest tests/test_phase1.py` or `uv run pytest -k <test_name>`.
+- **Run CLI in dev**: `uv run python -m blastradius.cli <command>` or `uv run blastradius <command>`.
+- **Environment note**: Avoid system `pip` or system `python` if broken; prefer `uv run` or `.venv\Scripts\` execution.
+
+## Core Architecture & Package Structure
+
+- **Package name**: `blastradius` (source in `blastradius/`). Entry point: `blastradius.cli:main`.
+- **Index artifacts**: Default dependency graph is written to `blastradius.json`; SQLite store is populated at `<repo>/.blastradius/index.db`.
+- **Layering rule**: `blastradius.store`, `blastradius.temporal`, and core modules must **never** import from `blastradius.semantic` or `blastradius.analyzers`.
+- **Backward compatibility**: The `blastradius.json` file export schema is frozen. SQLite DB synchronization is an additive side-effect of `index.build()`.
+
+## Storage & Database Gotchas (`blastradius/store/db.py`)
+
+- **FTS5 table**: Must remain a standalone table without the `content=` option. Using `content='symbols'` triggers `database disk image is malformed` errors during deletes in WAL mode.
+- **Soft deletes**: `sync()` must set `last_seen_commit` on soft-delete for temporal `--as-of` queries to filter correctly.
+- **SQLite transactions**: `executescript()` auto-commits pending transactions in Python `sqlite3`. Keep regular DDL and FTS DDL in separate `executescript()` calls.
+- **`sqlite-vec` optional dependency**: Never import `sqlite_vec` at module top-level. Wrap imports in `try...except` for runtime feature detection and graceful fallback to FTS5 + graph search.
+
+## Temporal Layer (`blastradius/temporal/history.py`)
+
+- **Git history reading**: `backfill()` streams commit history using `git ls-tree` and `git cat-file --batch`. Do not replace this with `git checkout` or individual `git show` calls.
+
+## Coding Conventions
+
+- **Module header**: Every Python file starts with:
+  ```python
+  # Copyright 2026 David Scheiderman
+  # Licensed under the Apache License, Version 2.0
+  from __future__ import annotations
+  ```
+- **Python version**: Targets Python 3.9+. Modern union (`X | Y`) and generic types (`list[str]`) are enabled via `from __future__ import annotations`.

CHANGELOG.md

@@ -6,7 +6,7 @@ All notable changes to this project will be documented in this file.
 
 ### Added
 
-- **`codeindex symbol-blast <file>`** — per-export blast radius for a file. Lists every
+- **`blastradius symbol-blast <file>`** — per-export blast radius for a file. Lists every
   exported symbol with a count and list of which importer files reference it by name.
   Answers "which routes use each export from lib/db/schema.ts" without manual grep.
   Supports `--json` for programmatic use.
@@ -46,10 +46,10 @@ All notable changes to this project will be documented in this file.
 
 ### Fixed
 
-- **`lookup` reads SQLite DB** — `codeindex lookup` and the `lookup_symbol` MCP tool
-  previously read `symbolindex.json`, which is only written by `codeindex symbols`.
+- **`lookup` reads SQLite DB** — `blastradius lookup` and the `lookup_symbol` MCP tool
+  previously read `symbolindex.json`, which is only written by `blastradius symbols`.
   Both now query `Store.lookup_by_name()` from the SQLite DB (the same source as
-  `codeindex search`), falling back to `symbolindex.json` only when no DB is present.
+  `blastradius search`), falling back to `symbolindex.json` only when no DB is present.
   Symbols found via search — including those extracted from destructured re-exports —
   are now consistently reachable via lookup.
 - **`lookup` output shows symbol name** — plain-mode output was `auth.ts:8 (const)`;
@@ -69,9 +69,9 @@ All notable changes to this project will be documented in this file.
   genuine API surfaces with broad real coupling.
 - **`changed-since` edge origin annotation** — added edges now carry `first_seen_commit`;
   when N edges share the `last_indexed_commit` value, the CLI prints a count. The message
-  branches on whether `codeindex history` has been run:
-  - History not run: `"run codeindex history to date them accurately"`
-  - History run: `"bootstrap-gap artifacts: existed before the first codeindex analyze
+  branches on whether `blastradius history` has been run:
+  - History not run: `"run blastradius history to date them accurately"`
+  - History run: `"bootstrap-gap artifacts: existed before the first blastradius analyze
     and cannot be dated further"` — correctly reflects the inherent limitation rather
     than implying a fixable error. MCP response gains `bootstrap_gap` boolean.
 
@@ -113,34 +113,34 @@ All notable changes to this project will be documented in this file.
   when FTS (or semantic KNN) already returns ≥ k results, preventing
   structurally adjacent but semantically unrelated symbols from diluting
   high-quality keyword hits.
-- **Search file aggregation** — `codeindex search` and the `semantic_search`
+- **Search file aggregation** — `blastradius search` and the `semantic_search`
   MCP tool now include a `Files` section aggregating results by file (sorted
   by symbol hit count). The entry-point file appears even when no single
   symbol from it ranks at the top.
-- **`db status` FTS row count** — `codeindex db status` now shows
+- **`db status` FTS row count** — `blastradius db status` now shows
   `fts_symbols` (rows in `symbols_fts`) making it easy to diagnose whether
   the FTS index is populated.
 
 ## [0.3.1] - 2026-06-07
 
 ### Fixed
 
-- **Multi-word FTS search OR fallback** — `codeindex search "auth token"` previously
+- **Multi-word FTS search OR fallback** — `blastradius search "auth token"` previously
   returned nothing when AND semantics found no single symbol containing all words.
   `fts_search()` now retries with `word1 OR word2 OR ...` automatically when the
   AND query returns zero results.
-- **`changed-since` backfill warning** — when `codeindex history` has never been run,
+- **`changed-since` backfill warning** — when `blastradius history` has never been run,
   all files share exactly one `first_seen_commit`, making `changed-since` results
   inaccurate against any older ref. The command now detects this via
   `COUNT(DISTINCT first_seen_commit) <= 1` and prints a clear warning on stderr
   (CLI) / includes a `"warning"` key in the response (JSON + MCP tool) directing
-  the user to run `codeindex history` first.
+  the user to run `blastradius history` first.
 
 ## [0.3.0] - 2026-06-07
 
 ### Summary
 
-codeindex evolves from a stateless point-in-time dependency analyzer into a
+blastradius evolves from a stateless point-in-time dependency analyzer into a
 **temporal code knowledge graph** — persistent, incremental, and semantically
 queryable. Three new properties: persistence + incrementality, time, and
 meaning (semantic retrieval). All existing CLI commands, JSON schemas, and MCP
@@ -149,36 +149,36 @@ tools are unchanged.
 ### Added
 
 #### Persistent SQLite store (Phase 1)
-- `.codeindex/index.db` — SQLite graph store created automatically on
-  `codeindex analyze`; survives across runs, never touches `codeindex.json`
+- `.blastradius/index.db` — SQLite graph store created automatically on
+  `blastradius analyze`; survives across runs, never touches `blastradius.json`
 - Incremental indexing: detects changed files via `git diff --name-status`
   between index runs; logs changed file count to stderr
-- `codeindex db status` — schema version, last indexed commit, file/edge/symbol counts
-- `codeindex db migrate` — applies pending schema migrations (runs automatically on open)
-- `codeindex symbols` now syncs symbols to DB with FTS5 full-text index
+- `blastradius db status` — schema version, last indexed commit, file/edge/symbol counts
+- `blastradius db migrate` — applies pending schema migrations (runs automatically on open)
+- `blastradius symbols` now syncs symbols to DB with FTS5 full-text index
 
 #### Temporal layer (Phase 2)
 - Every file, edge, and symbol carries `first_seen_commit` / `last_seen_commit`
   — facts are never hard-deleted, only soft-deleted with temporal stamps
-- `codeindex history [--since REF] [--max-commits N]` — backfills temporal
+- `blastradius history [--since REF] [--max-commits N]` — backfills temporal
   data from git history without any working-tree checkouts (uses
   `git ls-tree` + `git cat-file --batch`)
-- `codeindex changed-since <ref>` — files and edges added or removed since a
+- `blastradius changed-since <ref>` — files and edges added or removed since a
   commit, branch, or tag
-- `codeindex impact <file> --as-of <ref>` — blast radius at a historical point
+- `blastradius impact <file> --as-of <ref>` — blast radius at a historical point
   in time, not just HEAD
 
 #### Semantic layer (Phase 3)
-- `codeindex search "<query>" [--k N] [--as-of REF] [--json]` — hybrid
+- `blastradius search "<query>" [--k N] [--as-of REF] [--json]` — hybrid
   semantic + FTS5 keyword + graph expansion search, fused with Reciprocal Rank
   Fusion (RRF)
-- `codeindex/semantic/provider.py` — `EmbeddingProvider` ABC +
+- `blastradius/semantic/provider.py` — `EmbeddingProvider` ABC +
   `OpenAIEmbeddingProvider` HTTP client (stdlib `urllib` only, no new runtime deps)
 - `sqlite-vec` optional extension for KNN vector search; absent = graceful
   fallback to FTS + graph with a clear notice (no crash, no config required)
-- Embeddings generated automatically during `codeindex analyze` when
-  `CODEINDEX_EMBEDDING_ENDPOINT` / `_MODEL` / `_DIMS` env vars are set
-- `codeindex[semantic]` extra: `pip install 'codeindex[semantic]'`
+- Embeddings generated automatically during `blastradius analyze` when
+  `BLASTRADIUS_EMBEDDING_ENDPOINT` / `_MODEL` / `_DIMS` env vars are set
+- `blastradius[semantic]` extra: `pip install 'blastradius[semantic]'`
 
 #### MCP surface (Phase 4) — 4 new tools, existing 6 unchanged
 - `semantic_search` — hybrid search from an MCP client; degrades gracefully
@@ -190,28 +190,28 @@ tools are unchanged.
 - `schema_version` bumped to `"2"` with forward migration from `"1"`
 - FTS5 `symbols_fts` rowid now equals `symbols.id` (enables direct FTS → symbol
   row mapping without a secondary lookup)
-- `codeindex db status` output extended with `embedding_model`, `embedding_dims`,
+- `blastradius db status` output extended with `embedding_model`, `embedding_dims`,
   `vec_symbols` fields
 - README rewritten to document all new commands, the SQLite store, semantic
   setup, and the full 10-tool MCP surface
 
 ### Internal
-- `codeindex/store/db.py`: `Store` class — `init_vectors()`, `upsert_embeddings()`,
+- `blastradius/store/db.py`: `Store` class — `init_vectors()`, `upsert_embeddings()`,
   `semantic_search()`, `fts_search()`, `graph_expand()`, `neighborhood()`,
   `symbol_visible_at()`, `get_symbol()`, `symbols_needing_embeddings()`
-- `codeindex/temporal/history.py`: `backfill()` — BFS over git log via plumbing
+- `blastradius/temporal/history.py`: `backfill()` — BFS over git log via plumbing
   commands; no checkout side-effects
-- `codeindex/semantic/search.py`: `hybrid_search()` with RRF fusion
+- `blastradius/semantic/search.py`: `hybrid_search()` with RRF fusion
 - Dependency direction enforced: `store/` and `temporal/` never import from
   `semantic/` or `graph/`
 - 7 new Phase 3 tests; 6 Phase 2 tests; 5 Phase 1 tests (18 total, all green)
 
 ## [0.2.0] - 2026-05-24
 
 ### Added
-- `codeindex lookup <symbol>` — find where a symbol is defined (file + line)
-- `codeindex dependencies <file>` — show imports and imported-by for a file
-- `codeindex high-blast` — list files above a blast score threshold
+- `blastradius lookup <symbol>` — find where a symbol is defined (file + line)
+- `blastradius dependencies <file>` — show imports and imported-by for a file
+- `blastradius high-blast` — list files above a blast score threshold
 - All three new commands support `--json` for machine-readable output
 - `lookup_symbol` and `build_symbol_index` tools in MCP server
 - CLI integration test suite (`benchmark/test_cli.py`) — 37 assertions covering happy path, `--json` output, error cases, and sort-order invariants
@@ -231,12 +231,12 @@ tools are unchanged.
 ### Added
 - Multi-language dependency analysis: Python, JavaScript/TypeScript, Go, Ruby, Rust, Java/Kotlin, PHP, CSS
 - Blast-radius impact scoring — every file gets a score based on direct and transitive dependents
-- `codeindex analyze <repo>` — analyze a repo and write `codeindex.json`
-- `codeindex impact <file>` — show blast-radius impact report for a file
-- `codeindex symbols <repo>` — build `symbolindex.json` with functions, classes, and exports; supports `--inline` and `--claude-md` modes
-- `codeindex serve --mcp` — MCP stdio server exposing `analyze_repo`, `get_impact`, `get_dependencies`, `get_high_blast_files`, `build_symbol_index`, `lookup_symbol`
-- `codeindex serve --viz` — visualization UI server
-- `codeindex install-hook` — pre-commit hook for blast-radius warnings
+- `blastradius analyze <repo>` — analyze a repo and write `blastradius.json`
+- `blastradius impact <file>` — show blast-radius impact report for a file
+- `blastradius symbols <repo>` — build `symbolindex.json` with functions, classes, and exports; supports `--inline` and `--claude-md` modes
+- `blastradius serve --mcp` — MCP stdio server exposing `analyze_repo`, `get_impact`, `get_dependencies`, `get_high_blast_files`, `build_symbol_index`, `lookup_symbol`
+- `blastradius serve --viz` — visualization UI server
+- `blastradius install-hook` — pre-commit hook for blast-radius warnings
 - Phase 4: Docker, CI/CD, and schema analyzers
 - Phase 5: monorepo and cross-language intelligence
 - Apache 2.0 license