-- --- -- --:---- no signal
TUCK · last activity

Datahouse // Console

Enter the console password.

DATAHOUSE // CONSOLE
connecting poll — llm — version — snapshots/min —

Overview

Trend across versions

TUCK Datahouse Ingest

Human evidence, attached to a company and kept beside the audit — never inside it. Nothing here writes to a tracker row. Your words and an article's words are stored in separate fields and never merged, so a model reading this later can always tell which is which.

Paste a note

Add news

The article text is captured now and stored with the link. 43 links in this tracker are already dead; a citation that stops resolving is not a citation.

Batch analysis

A cohort is a rule, not a list of names, so the same batch can be re-run next month and diffed against this one. One LLM call per company, checkpointed after each — a restart resumes rather than starts over.

Search every note, thought and captured article

Timeline

Run the pipeline

Where your companies actually are

Adding a company writes a row to the ingest queue. The resolve stage turns that into an entity. Nothing in the package turns an entity into a tracker row — that link (graduation stage 2) was deferred on purpose so unverified claims cannot reach the export. This view exists so the missing link is visible rather than something you have to read the source to find.

Queue by status — and what reads it

A status with no consumer is terminal: rows in it will never move again on their own.

Drill in

Add companies

Upload a workbook


  

Versions

Runs

Entity lookup

Admin

code + config (secrets redacted) + a consistent DB backup

This session is a signed, HttpOnly cookie — no token is stored in this browser. Change the password on the server with sudo setdatahousepass; every session ends the moment you do.

Keyboard

/
focus the nearest filter
1…7
jump to a tab
g t
tuck ingest
g o
overview
g i
intake
g p
pipeline
g d
data
g e
entities
r
refresh now
t
cycle theme
?
this panel
Esc
close