Zero four zero

.github/workflows/ci.yml

@@ -26,21 +26,16 @@ jobs:
         run: |
           set -e
           BIN=./target/release/bitvanes
-          # --no-tui with no input prints help (and exits successfully).
-          $BIN --no-tui | head -1
-          # Markdown -> JSON / CSV / Arrow IPC
-          printf '# Hello\n\nThis is a test document.\n\nContact alice@example.com.\n' > /tmp/t.md
-          $BIN --no-tui -i /tmp/t.md -f markdown --scrub email -o /tmp/out.json
-          grep -q '"text"' /tmp/out.json
-          $BIN --no-tui -i /tmp/t.md -f markdown -o /tmp/out.csv
-          head -1 /tmp/out.csv | grep -q 'chunk_index'
-          $BIN --no-tui -i /tmp/t.md -f markdown -o /tmp/out.arrow
-          test -s /tmp/out.arrow
-          # JSON input
-          printf '{"name":"BitVanes","tags":["a","b"]}' > /tmp/t.json
-          $BIN --no-tui -i /tmp/t.json -o /tmp/outj.json
-          grep -q 'BitVanes' /tmp/outj.json
-          # Directory walk picks up all supported extensions
-          mkdir -p /tmp/docs && cp /tmp/t.md /tmp/docs/ && cp /tmp/t.json /tmp/docs/
-          $BIN --no-tui -i /tmp/docs/ -o /tmp/dir.json
-          grep -q '"chunk_index"' /tmp/dir.json
+          $BIN --no-tui | head -1                       # help when no input
+          mkdir -p /tmp/docs && printf '# Hello\n\nWorld.\n' > /tmp/docs/t.md
+          # markdown -> JSON / CSV / Arrow
+          $BIN --no-tui -i /tmp/docs/t.md -f markdown -o /tmp/o.json && grep -q text /tmp/o.json
+          $BIN --no-tui -i /tmp/docs/t.md -o /tmp/o.csv && grep -q chunk_index /tmp/o.csv
+          $BIN --no-tui -i /tmp/docs/t.md -o /tmp/o.arrow && test -s /tmp/o.arrow
+          # dedup keys in JSON
+          grep -q chunk_hash /tmp/o.json
+          # glob
+          $BIN --no-tui --glob "/tmp/docs/**/*.md" -o /tmp/g.json && grep -q chunk_index /tmp/g.json
+          # manifest idempotency
+          $BIN --no-tui -i /tmp/docs --manifest /tmp/m.json -o /tmp/r1.json
+          $BIN --no-tui -i /tmp/docs --manifest /tmp/m.json -o /tmp/r2.json 2>&1 | grep -q "Nothing new"

AGENTS.md

@@ -20,7 +20,8 @@ cat /tmp/out.json
 
 ## Architecture
 
-Headless CLI that links `bitvanes-core` via git dependency (tag `v0.1.0`).
+Headless CLI that links `bitvanes-core` via path dependency (local dev) or
+git tag `v0.3.0` (release).
 
 Entry point: `src/main.rs` (clap arg parsing → dispatch to headless or TUI).
 Headless: `src/headless.rs` (directory scan → rayon parallel → output).
@@ -34,19 +35,27 @@ Output formats: Arrow IPC, CSV, JSON (all from `bitvanes-core`'s `arrow_io`).
 ## Dependency on core
 
 ```toml
-bitvanes-core = { git = "https://github.com/BitVanes/core.git", tag = "v0.1.1", features = ["ipc", "csv", "cli-pdf", "parallel"] }
+bitvanes-core = { path = "../core/crates/core", features = ["ipc", "csv", "cli-pdf", "parallel"] }
+# For release: git = "https://github.com/BitVanes/core.git", tag = "v0.3.0"
 ```
 
 The CLI is distributed as a prebuilt binary via GitHub Releases (not published
 to crates.io), so it links core via a git tag. Only `bitvanes-core` is
 published to crates.io. To bump the core version, update the tag here and in
 Cargo.toml.
 
+## Features
+
+`default = []`. The `embed` cargo feature enables `--embed` (on-device
+embeddings via ONNX Runtime). It is **off by default** because linking the
+prebuilt `ort` static lib needs glibc ≥ 2.38, which the Linux release runner
+(ubuntu-22.04) lacks; build locally with `cargo build --features embed`.
+
 ## Release
 
 ```bash
-git tag v0.1.0
-git push origin v0.1.0
+git tag v0.3.0
+git push origin v0.3.0
 ```
 
 The release workflow (.github/workflows/release.yml) builds binaries for: