feat: self-reflection FP filtering, indicatif progress bars, 4 new languages

Cargo.toml

@@ -55,6 +55,10 @@ tree-sitter-python = "0.23"
 tree-sitter-typescript = "0.23"
 tree-sitter-javascript = "0.23"
 tree-sitter-go = "0.23"
+tree-sitter-java = "0.23"
+tree-sitter-c = "0.23"
+tree-sitter-cpp = "0.23"
+tree-sitter-ruby = "0.23"
 git2 = { version = "0.19", default-features = false }
 petgraph = "0.7"
 ignore = "0.4"
@@ -64,6 +68,7 @@ sha2 = "0.10"
 rmcp = { version = "0.15", features = ["server", "transport-io"] }
 miette = { version = "7", features = ["fancy"] }
 human-panic = "2"
+indicatif = "0.17"
 
 [package]
 name = "argus-ai"

PLAN.md

@@ -0,0 +1,57 @@
+# Argus Improvement Sprint — Plan
+
+## Feature 1: Self-Reflection / FP Filtering (HIGH PRIORITY)
+
+**Problem**: #1 user complaint is noise. AI reviews generate too many false positives — style nits, low-confidence hunches, and comments about code the LLM can't fully verify.
+
+**Solution**: After the initial LLM review generates comments, run a second LLM pass that evaluates each comment against the original diff. The second pass:
+1. Receives the original diff + all generated comments
+2. Scores each comment 1-10 for relevance and correctness
+3. Filters out comments scoring below threshold (default: 7)
+4. Can also re-classify severity (e.g., downgrade a "bug" to "suggestion")
+
+**Implementation**:
+- New function `self_reflect()` in `pipeline.rs`
+- New prompt `build_self_reflection_prompt()` in `prompt.rs`
+- New config field `self_reflection: bool` in `ReviewConfig` (default: true)
+- New config field `self_reflection_score_threshold: u8` (default: 7)
+- Stats tracked: `comments_reflected_out: usize`
+- Integrates between step 3 (deduplicate) and step 4 (filter_and_sort)
+- Uses same LlmClient — one additional API call
+
+**Research**: PR-Agent uses `new_score_mechanism` with thresholds. BugBot achieves 70%+ fix rate through aggressive filtering. Our approach: LLM-as-judge on its own output with structured scoring.
+
+## Feature 2: Progress Bars with indicatif (MEDIUM PRIORITY)
+
+**Problem**: Current progress is raw `eprintln!` text. Looks unprofessional and doesn't show elapsed time or spinners.
+
+**Solution**: Replace `eprintln!` progress with `indicatif` progress bars/spinners.
+- Spinner for single-group reviews
+- Multi-progress bar for multi-group reviews showing each group
+- Elapsed time display
+- Only when stderr is a terminal (already checked)
+
+**Implementation**:
+- Add `indicatif` dependency to `argus-review`
+- Refactor progress output in `pipeline.rs` review method
+- Use `ProgressBar` with spinner style for LLM calls
+- Use `MultiProgress` for multi-group reviews
+
+## Feature 3: More Tree-sitter Languages (LOW PRIORITY)
+
+**Problem**: Only supports Rust/Python/TypeScript/JavaScript/Go. Missing Java, C/C++, Ruby, PHP, Kotlin, Swift.
+
+**Solution**: Add tree-sitter grammars for Java, C, C++, Ruby. (PHP/Kotlin/Swift have less mature grammars.)
+
+**Implementation**:
+- Add `tree-sitter-java`, `tree-sitter-c`, `tree-sitter-cpp`, `tree-sitter-ruby` to workspace deps
+- Add language detection and parser setup in `parser.rs` and `walker.rs`
+- Add node type mappings for definitions/references per language
+- Tests for each new language
+
+## Execution Order
+1. Self-reflection (biggest impact)
+2. Progress bars (quick UX win)
+3. More languages (incremental)
+
+Each feature = one commit.