TEXXR CLI Reference

The TEXXR CLI (cli/texxr.py) queries the production API at texxr.com. All commands hit remote endpoints — no local database access required.

python3 cli/texxr.py <command> [options]

Signal Detection: `day`

The most powerful research command. Analyzes a single date and returns statistically scored signals.

python3 cli/texxr.py day 2026-02-09

Output sections:

CLUSTERS

Entity clusters with statistical scoring. Each entity that appears in 2+ articles gets:

Field	Meaning
`z_score`	Coverage anomaly vs 30-day baseline. z >= 3.0 = strong spike, z >= 2.0 = moderate spike, z <= -2.0 = unusual silence
`source_count`	Number of distinct publications covering this entity
`convergence_score`	`articles × (distinct_sources / articles)`. High = many outlets converging on the story
`SPIKE`	z >= 3.0 — entity is getting far more coverage than its 30-day average
`NOVEL`	Entity has no 30-day baseline (first appearance or very rare)
`~`	Provisional — fewer than 7 days of baseline data

How to read it: Sort by z-score. The top z-scores are entities experiencing statistically unusual coverage. A high z-score + high convergence_score = many outlets covering an entity that doesn't usually get this much attention. That's almost always a signal.

SILENCES

Entities that are active in the 30-day window but absent today. Sorted by expected daily rate.

How to read it: An entity averaging 2+ articles/day that goes to zero is notable. Ask: is there a reason they're quiet? Is everyone else talking about something that affects them?

CO-OCCURRENCE SPIKES

Entity pairs appearing together more than their 90-day baseline predicts. Uses Poisson variance for sparse count data.

Field	Meaning
`today`	Co-occurrences today
`daily_rate`	Average co-occurrences per day over 90 days
`z_score`	Anomaly score (z >= 2.0 shown)
`NOVEL`	Pair has never co-occurred in 90 days

How to read it: Novel co-occurrences between major entities are often the most interesting signal. If Anthropic and OpenAI suddenly appear together at z=4.9 when their baseline is 0.14/day, something is connecting them today.

EDGE PROFILE

Distribution of relationship types from the knowledge graph for today's articles. Uses canonical labels: financial, launch, controversy, regulatory, personnel, legal, partnership, acquisition, competitor.

How to read it: Unusual distributions suggest the day's news has a particular character. A day heavy on controversy + legal is different from one heavy on launch + financial.

JUXTAPOSITIONS

Entities appearing in contradictory relationship types on the same day (e.g., launch + controversy, partnership + competitor). These are often the seed of a constellation.

How to read it: An entity in both financial and controversy edges on the same day has a tension worth investigating.

Research Workflow

Starting a day's research

# 1. Get the signal overview
python3 cli/texxr.py day 2026-02-09

# 2. Read the top articles
python3 cli/texxr.py articles --date=2026-02-09 --limit=50

# 3. Investigate the top z-score entity
python3 cli/texxr.py edges "Anthropic" -v --limit=20
python3 cli/texxr.py entity "Anthropic" --limit=20

Investigating a signal

# A co-occurrence spike between A and B — what connects them?
python3 cli/texxr.py edges-between "Anthropic" "OpenAI" -v

# A silence — what was this entity doing recently?
python3 cli/texxr.py entity "Microsoft" --limit=15
python3 cli/texxr.py timeline "Microsoft" --since=2026-01-15 --edges

# A juxtaposition — what's the tension?
python3 cli/texxr.py edges "OpenAI" -v --limit=30

Building historical context

# How has an entity's coverage evolved?
python3 cli/texxr.py momentum "OpenAI" --year=2026
python3 cli/texxr.py arc "AI regulation"

# Find similar stories across time
python3 cli/texxr.py similar <article_id> --limit=15

# Semantic search for a concept
python3 cli/texxr.py triangulate "AI companies selling advertising"

# What happened around a specific article?
python3 cli/texxr.py context <article_id> --window=5

All Commands

Browsing

Command	Usage	Purpose
`today`	`today --limit=35`	Today's articles sorted by position
`articles`	`articles --date=2026-02-09 --limit=50`	Articles for a specific date
`article`	`article 1162353` or `article 1162353,1162354`	Full details for specific article(s)

Search

Command	Usage	Purpose
`search`	`search "DeepSeek" --limit=10`	Text search across headlines and descriptions
`entity`	`entity "OpenAI" --limit=30`	Articles mentioning a specific entity
`similar`	`similar 1162353 --limit=15`	Vector similarity search from an article
`triangulate`	`triangulate "AI companies selling ads"`	Semantic search (embedding-based)

Knowledge Graph

Command	Usage	Purpose
`edges`	`edges "OpenAI" -v --limit=20`	Relationship edges involving an entity
`edges-between`	`edges-between "Google" "Apple" -v`	Edges connecting two specific entities
`article-edges`	`article-edges 1162353`	All edges extracted from a specific article
`edge-profile`	`edge-profile "OpenAI" --since=2025-01-01`	Predicate distribution (narrative fingerprint)
`edge-search`	`edge-search "company launches ads" --limit=10`	Semantic search on edge summaries

Arc Discovery

Command	Usage	Purpose
`edge-arc`	`edge-arc "Microsoft" --granularity=quarter`	Predicate evolution over time with centroid drift and co-entity shifts
`regime`	`regime "Microsoft" --threshold=0.25`	Multi-dimensional break detection (JSD + semantic drift + network Jaccard)
`thread`	`thread "Microsoft" --with=Anthropic`	Chronological edge timeline with event flags (FIRST_EDGE, STATUS_CHANGE, CONVERGENCE, ENTERED)
`status-flow`	`status-flow --entity="OpenAI" -v`	Edge status lifecycle: distribution, long-pending, confirmed, failed/denied
`web`	`web Microsoft OpenAI Anthropic`	Entity constellation: connections, tensions, disconnected entities

Analysis

Command	Usage	Purpose
`day`	`day 2026-02-09`	Signal detection — z-scores, silences, co-occurrence spikes, edge profile, juxtapositions
`momentum`	`momentum "OpenAI" --year=2026`	Coverage velocity and acceleration
`arc`	`arc "AI regulation"`	Topic evolution over time
`discover`	`discover`	Auto-discover patterns (drift, correlations, collisions, silences)
`timeline`	`timeline "xAI" --since=2024-06-01 --edges`	Chronological articles + edge phases
`compare`	`compare 2026-01-15 2026-01-20 --entity "OpenAI"`	Coverage comparison between two dates
`context`	`context 1162353 --window=5`	Articles before/after a specific article
`sources`	`sources "AI safety" --date-range=2025-01-01:2025-12-31`	Which outlets cover a topic

Flags

Flag	Effect
`-c` / `--compact`	One line per article (saves context window)
`-v` / `--verbose`	Show edge summaries and descriptions
`--limit=N`	Limit number of results
`--date-range=START:END`	Filter by date range (YYYY-MM-DD:YYYY-MM-DD)

Tips for Agents

Start with day when researching a specific date. It surfaces the statistically interesting signals so you don't have to manually scan articles.
Use -c (compact) for browsing commands (today, articles, search, entity). It saves context window by printing one line per article.
Use -v (verbose) for edge commands. Without it, edge summaries are hidden and you only see predicate/entity names.
Article IDs are sequential integers. Jan 2026 starts around 1154xxx. You can batch-lookup with comma-separated IDs: article 1162353,1162354,1162355 -c.
triangulate is the most powerful search — it uses vector embeddings, not keywords. Use it when keyword search misses semantically related articles.
edges-between is essential for investigating co-occurrence spikes. It shows the actual relationship edges between two entities.
edge-profile gives an entity's "narrative fingerprint" — the distribution of relationship types. An entity that's 40% controversy + 30% legal has a very different profile from one that's 50% launch + 30% financial.
Z-score interpretation:
z >= 3.0: Almost certainly a real signal — investigate
z 2.0-3.0: Probably real, worth a quick look
z 1.0-2.0: Could be noise
z <= -2.0: Notable absence — why is this entity quiet?