The survey pass no longer receives the bucketed file_categories
histogram, which was biased toward source-code targets and would
mislabel mail, notebooks, ledgers, and other non-code domains as
"source" via the file --brief "text" pattern fallback.
Adds filetypes.survey_signals(), which assembles raw signals from
the same `classified` data the bucketer already processes — no new
walks, no new dependencies:
total_files — total count
extension_histogram — top 20 extensions, raw, no taxonomy
file_descriptions — top 20 `file --brief` outputs, by count
filename_samples — 20 names, evenly drawn (not first-20)
`survey --brief` descriptions are truncated at 80 chars before
counting so prefixes group correctly without exploding key cardinality.
The Band-Aid in _SURVEY_SYSTEM_PROMPT (warning the LLM that the
histogram was biased toward source code) is removed and replaced
with neutral guidance on how to read the raw signals together.
The {file_type_distribution} placeholder is renamed to
{survey_signals} to reflect the broader content.
luminos.py base scan computes survey_signals once and stores it on
report["survey_signals"]; AI consumers read from there.
summarize_categories() and report["file_categories"] are unchanged
— the terminal report still uses the bucketed view (#49 tracks
fixing that follow-up).
Smoke tested on two targets:
- luminos_lib: identical-quality survey ("Python library package",
confidence 0.85), unchanged behavior on code targets.
- A synthetic Maildir of 8 messages with `:2,S` flag suffixes:
survey now correctly identifies it as "A Maildir-format mailbox
containing 8 email messages" with confidence 0.90, names the
Maildir naming convention in domain_notes, and correctly marks
parse_structure as a skip tool. Before #42 this would have been
"8 source files."
Adds 8 unit tests for survey_signals covering empty input, extension
histogram, description aggregation/truncation, top-N cap, and
even-stride filename sampling.
#48 tracks the unit-of-analysis limitation (file is the wrong unit
for mbox, SQLite, archives, notebooks) — explicitly out of scope
for #42 and documented in survey_signals' docstring.
Adds a gate in _run_investigation that skips the survey API call when
a target has both fewer than _SURVEY_MIN_FILES (5) files AND fewer
than _SURVEY_MIN_DIRS (2) directories. AND semantics handle the
deep-narrow edge case correctly: a target with 4 files spread across
50 directories still gets a survey because dir count amortizes the
cost across 50 dir loops.
When skipped, _default_survey() supplies a synthetic dict with
confidence=0.0 — chosen specifically so _filter_dir_tools() never
enforces skip_tools from a synthetic value. The dir loop receives
a generic "small target, read everything" framing in its prompt and
keeps its full toolbox.
Reorders _discover_directories() to run before the survey gate so
total_dirs is available without a second walk.
#46 tracks revisiting the threshold values with empirical data after
Phase 2 ships and we've run --ai on a variety of real targets.
Smoke tested on a 2-file target: gate triggers, default survey
substituted, dir loop completes normally. Adds 4 unit tests for
_default_survey() covering schema, confidence guard, filter
interaction, and empty skip_tools.
The survey pass now actually steers dir loop behavior, in two ways:
1. Prompt injection: a new {survey_context} placeholder in
_DIR_SYSTEM_PROMPT receives the survey description, approach,
domain_notes, relevant_tools, and skip_tools so the dir-loop agent
has investigation context before its first turn.
2. Tool schema filtering: _filter_dir_tools() removes any tool listed
in skip_tools from the schema passed to the API, gated on
survey confidence >= 0.5. Control-flow tools (submit_report) are
always preserved. This is hard enforcement — the agent literally
cannot call a filtered tool, which the smoke test for #5 showed
was necessary (prompt-only guidance was ignored).
Smoke test on luminos_lib: zero run_command invocations (vs 2 before),
context budget no longer exhausted (87k vs 133k), cost ~$0.34 (vs
$0.46), investigation completes instead of early-exiting.
Adds tests/test_ai_filter.py with 14 tests covering _filter_dir_tools
and _format_survey_block — both pure helpers, no live API needed.
Adds the reconnaissance survey pass: a fast, ≤3-turn LLM call that
characterizes the target before any directory investigation begins.
The survey receives the file-type distribution (from the base scan),
a top-2-level tree preview, and the list of available dir-loop tools,
and returns description / approach / relevant_tools / skip_tools /
domain_notes / confidence via a single submit_survey tool call.
Wired into _run_investigation() before the directory loop. Output is
logged but not yet consumed — that wiring is #6. Survey failure is
non-fatal: if the call errors or runs out of turns, the investigation
proceeds without survey context.
Also adds a Band-Aid to _SURVEY_SYSTEM_PROMPT warning the LLM that
the file-type histogram is biased toward source code (the underlying
classifier has no concept of mail, notebooks, ledgers, etc.) and to
trust the tree preview when they conflict. The proper fix is #42.
Moves _DIR_SYSTEM_PROMPT and _SYNTHESIS_SYSTEM_PROMPT from ai.py into
a dedicated prompts module. Both are pure template strings with .format()
placeholders — no runtime imports needed in prompts.py. Prompt content
is byte-for-byte identical to the original.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Moves all tree-sitter parsing logic from ai.py into a dedicated module.
Replaces the if/elif language chain with a _LANGUAGE_HANDLERS registry
mapping language names to handler functions.
Extracted: _tool_parse_structure body, _get_ts_parser, _child_by_type,
_text, and all per-language helpers (_py_func_sig, _py_class, etc.).
ai.py retains a thin wrapper for path validation.
Public API: parse_structure(path) -> JSON string
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Delete unused clear_cache() from ai.py (luminos.py imports it from
capabilities.py)
- Remove CACHE_ROOT import from ai.py (was only used by dead function)
- Replace local CACHE_ROOT constant in capabilities.py with import
from cache.py (single source of truth)
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Moves investigation ID persistence and _CacheManager class from ai.py
into a dedicated cache module. No behavior changes.
Moved: _load_investigations, _save_investigations, _get_investigation_id,
_CacheManager (all methods), _sha256_path, CACHE_ROOT, INVESTIGATIONS_PATH.
Also added a local _now_iso() in cache.py to avoid a circular import
(ai.py imports from cache.py).
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Adds think, checkpoint, and flag tools for agent reasoning visibility:
- think: records observation/hypothesis/next_action before investigation
- checkpoint: summarizes learned/unknown/next_phase after file clusters
- flag: marks notable findings to flags.jsonl with severity levels
Additional changes:
- Step numbering in investigation system prompt
- Text blocks from agent now printed to stderr (step labels visible)
- flag tool available in both investigation and synthesis passes
- analyze_directory() returns (brief, detailed, flags) three-tuple
- format_flags() in report.py renders flags sorted by severity
- Per-directory max_turns increased from 10 to 14
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
When the 70% context budget is hit mid-directory, the early exit now
writes a partial directory cache entry from whatever file summaries
the agent cached in prior turns, instead of discarding the work.
If file entries exist: concatenates their summaries into a directory
entry marked partial=true. If no files were cached: writes a minimal
entry noting the budget was reached before processing.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Rewrites ai.py from a single Claude API call into a multi-pass,
cache-driven agent architecture:
- Per-directory isolated agent loops (max 10 turns each) with context
discarded between directories
- Leaves-first processing order so child summaries inform parents
- Disk cache (/tmp/luminos/{uuid}/) persists across runs for resumability
- Investigation ID persistence keyed by target realpath
- Separate synthesis pass reads only directory-level cache entries
- Replaces urllib with Anthropic SDK (streaming, automatic retries)
- Token counting with 70% context budget threshold for early exit
- parse_structure tool via tree-sitter (Python, JS, Rust, Go)
- python-magic integration for MIME-aware directory listings
- Cost tracking printed at end of investigation
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Adds --ai flag that sends the directory tree, file categories, and
sampled file contents to Claude for analysis. Produces a brief
summary at the top of the report and a detailed breakdown at the
end. Requires ANTHROPIC_API_KEY env var; degrades gracefully without it.
Uses only stdlib (urllib) to keep the zero-dependency constraint.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
