neuroskill-screenshots

name

neuroskill-screenshots

description

NeuroSkill screenshot search commands — `search-images` for OCR text and CLIP vision search, `screenshots-around` for temporal proximity, `screenshots-for-eeg` for finding what was on screen during EEG sessions, and `eeg-for-screenshots` for discovering brain state near matching screenshots. Cross-modal EEG↔screenshot bridging. Use when searching for past screen content, correlating screen activity with brain states, or exploring screenshot history.

NeuroSkill Screenshot Search

NeuroSkill can periodically capture screenshots of the active window, embed them with CLIP vision (ONNX), run OCR text extraction, and index both modalities in separate HNSW indices for similarity search.

Note: Screenshot capture must be enabled in Settings → Screenshots. Screenshots are served via GET /screenshots/<filename>?token=<token> on the daemon API (auth required) and appear as indicators on the History day-view heatmap grid.

---

LLM Tool Calls

When calling these commands via the LLM skill tool, use command + args:

{"command": "search_screenshots", "args": {"query": "browser", "k": 10}}
{"command": "search_screenshots", "args": {"query": "TODO", "mode": "substring"}}
{"command": "screenshots_around", "args": {"timestamp": 1740412800, "window_secs": 60}}
{"command": "screenshots_for_eeg", "args": {"start_utc": 1740412800, "end_utc": 1740415500}}
{"command": "eeg_for_screenshots", "args": {"query": "compiler error", "k": 5}}

---

Command Overview

Command	Direction	Description
search-images "query"	Text → Screenshots	Search by OCR text (semantic/substring)
search-images --by-image <path>	Image → Screenshots	Search by visual similarity (CLIP)
screenshots-around --at <utc>	Timestamp → Screenshots	Find screenshots near a specific time
screenshots-for-eeg	EEG → Screenshots	Find screenshots captured during EEG epochs
eeg-for-screenshots "query"	Screenshots → EEG	Find brain data & labels near matched screenshots

---

search-images — Search Screenshots by OCR Text or Visual Similarity

Search all captured screenshots using OCR-extracted text or CLIP vision embeddings.

OCR Text Search Modes

• semantic (default) — embeds the query text and searches the OCR HNSW index
• substring — direct SQL LIKE matching against raw OCR text

npx neuroskill search-images "compiler error"
npx neuroskill search-images "login page" --k 5
npx neuroskill search-images "TODO" --mode substring
npx neuroskill search-images "dashboard" --json | jq '.results[].filename'
npx neuroskill search-images "slack message" --mode semantic --k 10
npx neuroskill search-images "code review" --json | jq '.results[] | {file: .filename, app: .app_name, sim: .similarity}'

Vision Search (CLIP)

Search by visual similarity using a reference image. The image is embedded via the CLIP model and compared against the screenshots HNSW index.

npx neuroskill search-images --by-image reference.png
npx neuroskill search-images --by-image screenshot.jpg --k 5
npx neuroskill search-images --by-image layout-mock.webp --json | jq '.results[].filename'

Options

Flag	Default	Description
--mode <m>	semantic	Search mode: semantic (embedding similarity) or substring (literal text match)
--by-image <path>	—	Search by visual similarity instead of OCR text (CLIP vision)
--k <n>	20	Maximum number of results to return
--json	—	Output raw JSON (pipe-safe)
--full	—	Human-readable summary + colorized JSON

HTTP / WebSocket API

# Semantic OCR search (default):
curl -s -X POST http://127.0.0.1:8375/ \
  -H "Content-Type: application/json" \
  -d '{"command":"search_screenshots","query":"compiler error","k":20,"mode":"semantic"}'

# Substring OCR search:
curl -s -X POST http://127.0.0.1:8375/ \
  -H "Content-Type: application/json" \
  -d '{"command":"search_screenshots","query":"TODO","k":10,"mode":"substring"}'

# Vision search (base64-encoded image):
IMAGE_B64=$(base64 -w0 reference.png)
curl -s -X POST http://127.0.0.1:8375/ \
  -H "Content-Type: application/json" \
  -d "{\"command\":\"search_screenshots_by_image_b64\",\"image_b64\":\"$IMAGE_B64\",\"k\":20}"

# Vision search (raw embedding vector):
curl -s -X POST http://127.0.0.1:8375/ \
  -H "Content-Type: application/json" \
  -d '{"command":"search_screenshots_vision","vector":[0.1, 0.2, ...],"k":20}'

JSON Response

{
  "command": "search_screenshots",
  "ok": true,
  "query": "compiler error",
  "mode": "semantic",   // or "vision" for --by-image
  "k": 20,
  "count": 3,
  "results": [
    {
      "timestamp":    20260315143025,
      "unix_ts":      1741963825,
      "filename":     "20260315/20260315143025.webp",
      "app_name":     "VS Code",
      "window_title": "skill — cli.ts",
      "ocr_text":     "error[E0308]: mismatched types expected…",
      "similarity":   0.87
    }
  ]
}

---

screenshots-around — Find Screenshots Near a Timestamp

Find all screenshots captured within a time window (±seconds) around a specific Unix timestamp. Useful for seeing what was on screen during a specific EEG moment.

npx neuroskill screenshots-around --at 1740412800
npx neuroskill screenshots-around --at 1740412800 --seconds 120
npx neuroskill screenshots-around --at 1740412800 --json | jq '.results[].filename'

Options

Flag	Default	Description
--at <utc>	(required)	Target Unix timestamp in seconds
--seconds <n>	60	Window size: ±N seconds around the target

HTTP / WebSocket API

curl -s -X POST http://127.0.0.1:8375/ \
  -H "Content-Type: application/json" \
  -d '{"command":"screenshots_around","timestamp":1740412800,"window_secs":60}'

JSON Response

{
  "command": "screenshots_around",
  "ok": true,
  "count": 4,
  "results": [
    {
      "timestamp":    20260224075945,
      "unix_ts":      1740412785,
      "filename":     "20260224/20260224075945.webp",
      "app_name":     "Firefox",
      "window_title": "GitHub - Pull Request",
      "ocr_text":     "Merge pull request #42..."
    }
  ]
}

---

screenshots-for-eeg — EEG → Screenshots

Find screenshots captured near EEG recording timestamps. Given an EEG time range (auto: latest session, or --start/--end), finds all EEG embedding timestamps in that range, then returns screenshots within --window seconds (default ±30s) of each epoch.

This is the "EEG → screenshots" cross-modal bridge: start from brain data, discover what was on screen at that moment.

npx neuroskill screenshots-for-eeg                             # auto: last session
npx neuroskill screenshots-for-eeg --start 1740412800 --end 1740415500
npx neuroskill screenshots-for-eeg --window 60                  # ±60s around each epoch
npx neuroskill screenshots-for-eeg --limit 20                   # max 20 results
npx neuroskill screenshots-for-eeg --json | jq '.results[].screenshot.filename'
npx neuroskill screenshots-for-eeg --json | jq '.results[] | {eeg: .eeg_timestamp_utc, file: .screenshot.filename}'

Options

Flag	Default	Description
--start <utc>	auto (latest session)	Start of EEG range (unix seconds)
--end <utc>	auto (latest session)	End of EEG range (unix seconds)
--window <n>	30	Temporal window: ±N seconds around each EEG epoch
--limit <n>	50	Maximum number of screenshot results

HTTP / WebSocket API

curl -s -X POST http://127.0.0.1:8375/ \
  -H "Content-Type: application/json" \
  -d '{"command":"screenshots_for_eeg","start_utc":1740412800,"end_utc":1740415500,"window_secs":30,"limit":50}'

JSON Response

{
  "command": "screenshots_for_eeg",
  "ok": true,
  "start_utc":   1740412800,
  "end_utc":     1740415500,
  "window_secs": 30,
  "eeg_count":   541,
  "count":       12,
  "results": [
    {
      "eeg_timestamp_utc": 1740413130,
      "screenshot": {
        "timestamp":    20260224080530,
        "unix_ts":      1740413128,
        "filename":     "20260224/20260224080530.webp",
        "app_name":     "VS Code",
        "window_title": "main.rs",
        "ocr_text":     "fn dispatch(app: &AppHandle…",
        "similarity":   0.0
      }
    }
  ]
}

---

eeg-for-screenshots — Screenshots → EEG

Find EEG data and labels near screenshots matching an OCR text query. This is the "screenshots → EEG" cross-modal bridge: start from what was on screen, discover what your brain was doing at that moment.

Pipeline:

1. Search screenshots by OCR text (semantic or substring mode).
2. For each matched screenshot, find EEG labels within --window seconds.
3. Report the associated EEG session and labels alongside each screenshot.

Use this to answer: "When I was looking at [X], what was my brain doing?"

npx neuroskill eeg-for-screenshots "compiler error"
npx neuroskill eeg-for-screenshots "TODO" --mode substring --k 5
npx neuroskill eeg-for-screenshots "dashboard" --window 120
npx neuroskill eeg-for-screenshots "login page" --json | jq '.results[].labels[].text'
npx neuroskill eeg-for-screenshots "slack" --json | jq '.results[] | {file: .screenshot.filename, labels: [.labels[].text]}'

Options

Flag	Default	Description
--mode <m>	semantic	Search mode: semantic or substring
--k <n>	10	Maximum number of screenshot matches
--window <n>	60	Temporal window: ±N seconds to search for EEG labels

HTTP / WebSocket API

curl -s -X POST http://127.0.0.1:8375/ \
  -H "Content-Type: application/json" \
  -d '{"command":"eeg_for_screenshots","query":"compiler error","k":10,"window_secs":60,"mode":"semantic"}'

JSON Response

{
  "command": "eeg_for_screenshots",
  "ok": true,
  "query":            "compiler error",
  "mode":             "semantic",
  "k":                10,
  "window_secs":      60,
  "screenshot_count": 3,
  "count":            3,
  "results": [
    {
      "screenshot": {
        "timestamp":    20260315143025,
        "unix_ts":      1741963825,
        "filename":     "20260315/20260315143025.webp",
        "app_name":     "VS Code",
        "window_title": "skill — cli.ts",
        "ocr_text":     "error[E0308]: mismatched types expected…",
        "similarity":   0.87
      },
      "labels": [
        {
          "id": 47, "text": "debugging session",
          "eeg_start": 1741963560, "eeg_end": 1741963860,
          "label_start": 1741963680, "label_end": 1741963680
        },
        {
          "id": 48, "text": "frustrated",
          "eeg_start": 1741963740, "eeg_end": 1741964040,
          "label_start": 1741963860, "label_end": 1741963860
        }
      ],
      "session": {
        "csv_path": "/home/user/.skill/20260315/exg_1741960800.csv",
        "session_start_utc": 1741960800,
        "session_end_utc":   1741964400,
        "device_name":       "Muse-A1B2"
      }
    }
  ]
}

---

Cross-Modal Query Guide

Choose the right command based on the direction of your query:

Question pattern	Direction	Command	Example
What apps did I use today?	—	status	Check apps.top_24h in the response
Find screenshots showing X	Text → Screenshots	search_screenshots	{"command":"search_screenshots","args":{"query":"VS Code"}}
What was on screen during my EEG session?	EEG → Screenshots	screenshots_for_eeg	{"command":"screenshots_for_eeg","args":{"start_utc":T1,"end_utc":T2}}
How was my brain when I saw X?	Screenshots → EEG	eeg_for_screenshots	{"command":"eeg_for_screenshots","args":{"query":"error message"}}
What was on screen at timestamp T?	Timestamp → Screenshots	screenshots_around	{"command":"screenshots_around","args":{"timestamp":T}}
Find brain states similar to "focus"	Text → Labels → EEG	interactive_search	{"command":"interactive_search","args":{"query":"deep focus"}}

Multi-Step Cross-Modal LLM Workflows

Chain commands for deeper cross-modal analysis:

"What was on screen during my best focus session?"

1. {"command": "sessions"} → get session timestamps
2. {"command": "screenshots_for_eeg", "args": {"start_utc": <start>, "end_utc": <end>, "window_secs": 15}}

"How did my brain react to error messages?"

1. {"command": "eeg_for_screenshots", "args": {"query": "error", "window_secs": 120}} → returns screenshots with matching EEG labels and session data

"What was I looking at when I labelled 'deep focus'?"

1. {"command": "search_labels", "args": {"query": "deep focus"}} → get created_at timestamp
2. {"command": "screenshots_around", "args": {"timestamp": <created_at>, "window_secs": 30}}

"Find similar brain states to my coding session and show what was on screen"

1. {"command": "search", "args": {"start_utc": <start>, "end_utc": <end>}} → get neighbor timestamps
2. {"command": "screenshots_around", "args": {"timestamp": <neighbor_ts>}} for each result

CLI Workflows

# ── "What was on screen during my best focus session?" ────────────────────────
START=$(npx neuroskill sessions --json | jq '.sessions[0].start_utc')
END=$(npx neuroskill sessions --json | jq '.sessions[0].end_utc')
npx neuroskill screenshots-for-eeg --start $START --end $END --window 15

# ── "How did my brain react to error messages?" ──────────────────────────────
npx neuroskill eeg-for-screenshots "error" --window 120

# ── "Find screenshots visually similar to a reference layout" ─────────────────
npx neuroskill search-images --by-image reference-layout.png --k 10

# ── "Find what I was looking at when a specific label was created" ────────────
LABEL_TS=$(npx neuroskill search-labels "deep focus" --json | jq '.results[0].created_at')
npx neuroskill screenshots-around --at $LABEL_TS --seconds 30

# ── "Full chain: OCR → screenshots → EEG → labels → neural search" ───────────
npx neuroskill eeg-for-screenshots "pull request" --json | jq '.results[0].session'
npx neuroskill search --start <session_start> --end <session_end> --k 5

# ── "See what was on screen at a search hit timestamp" ────────────────────────
HIT_TS=$(npx neuroskill search --json | jq '.result.results[0].neighbors[0].timestamp_unix')
npx neuroskill screenshots-around --at $HIT_TS

# ── "View the actual screenshot image in a browser" ──────────────────────────
PORT=8375
FILE=$(npx neuroskill screenshots-around --at 1740412800 --json | jq -r '.results[0].filename')
xdg-open "http://127.0.0.1:$PORT/screenshots/$FILE?token=$TOKEN"