The business caseThe problem this solves

Verifying a bank or brokerage statement for a loan file means reading the full deposit list, checking whether the period's largest deposit is documented well enough to skip a source-of-funds request, and computing a reserve value from the account's ending balance and type -- for every statement in the file. Someone manually opening each bank or brokerage statement, reading the deposit list, and deciding by eye whether the period's largest deposit needs a source-of-funds explanation before it goes in a loan file.

The inputThe actual statements

The corpus is 55 statements, 0.03 MB (txt 55). Plain text, one format, invented rather than fetched — a real bank or brokerage statement cannot be published at all (see SOURCES.md). Ten fields are chosen because the same ones matter to underwriting asset verification: who, where, what kind of account, the two balances, and whether the period's largest deposit has a documented source.

Statement Holder
----------------
Amara W. Haddad

Institution
-----------
Beacon Point Bank

Account Type
------------
savings

Statement Period
----------------
2026-06-01 to 2026-06-30

Account Summary
---------------
Beginning Balance: $18941.17
Ending Balance: $18099.20

Deposits This Period
--------------------
2026-06-07   $133.03   Mobile Check Deposit
2026-06-10   $168.01   ATM Deposit
2026-06-15   $238.78   Zelle from Amara N.
2026-06-19   $185.39   Mobile Check Deposit

Other Activity
--------------
Debits and fees this period: $1567.18

The outcomeWhat a good result looks like

A ten-field extracted record per statement, plus two pure-code computed figures -- a vested reserve value and whether the period's largest deposit needs underwriter review -- informational only, never an automatic approval or denial.

Where it fitsWhat did work

Every line below is a measured result from this kit's own runs, with the figure that supports it. The headline above is not softened by any of them.

• Screening a stack of bank/brokerage statements for which ones need a human to chase a source-of-funds letter before a loan file closes — either tier — they tie on every accuracy figure measured here
  100% flag recall and precision on both tiers this run, against the free keyword floor's 88.9% recall (1 missed large, undocumented deposit).
• Deciding the two tiers on cost or speed when accuracy is identical — the fast tier
  Roughly half the p50 latency (2,490 ms vs 5,064 ms) and 9% fewer output tokens per call for identical scores on this corpus.

At a glanceHow the whole thing runs

Run once, for real, on 2026-08-20. Every figure on these pages was captured from that run — nothing is written from intent.

14 steps, grouped by the question that sends you to them rather than by build order. Each tile carries the one figure that step is about, and opens the page behind it.

Should you use this?What you bring, where it stops, and when not to use it

Before you commit an afternoon to this, these are the answers that decide it. Each one is rendered from the record it lives in — and links the page that holds it in full.

Not asked of this kit — 2 questions: clone (a fresh clone of this kit runs with nothing fetched); judge (nothing here is graded by a model).

Last verified 2026-08-20 — r001-asset-verify. Every figure on these pages was captured from that run.

Run itHow this reaches your data

Every result on this page was produced by pure code over checked-in files, with no API key — which is why you can read the numbers before anyone spends anything.

Checked before this shipped — Verified from a cold clone with no key configured: 55 statements segment into 385 sections in 0.003 seconds, and the assembled prompt for AV-0001 replays byte for byte — its three part sizes (1163/1204/484 characters) match what run r001 recorded.

Step 01 of 14Business case

Written for: product manager · Measured during the eval run. Never estimated, never quoted from a vendor page.

In brief

What the clock covers. model call only, one per statement

Current processWhat it replaces

Someone manually opening each bank or brokerage statement, reading the deposit list, and deciding by eye whether the period's largest deposit needs a source-of-funds explanation before it goes in a loan file.

Step 02 of 14Architecture

Written for: solutions architect · Component map derived from the real code, not drawn from intent.

In brief

CorpusCorpusArchitectureArchitecturePromptPromptPromptPromptApp UIApp UIEvalsEvalsBusiness caseUnit cost

• 1stations numbered in run order
• txt
• 1 of 7 stations carry a recorded failure

scroll → 7 stations

One question, end to end. Colour on the fan-in is the corpus format; a red hook marks every station with a recorded failure, and the words — every sub line, failure and report field — are in the ledger beneath. Every station is a link — click one to jump to its row, and each row links to the lens that explains it.

The swap seams

Components

Step 03 of 14App UI

Written for: anyone, in 5 seconds · Screenshots of the running app taken during the run.

SuccessesWhen it works

LimitsWhen it does not

A report showing only wins is an advert. This one is required, and the validator now refuses a kit that omits it.

Step 04 of 14Corpus

Written for: "can I point this at mine?" · Measured off the corpus the run actually used.

In brief

How it is cutWhat one section is

cut on underlined section headings; a statement with none falls back to one whole-document segment so a span still resolves

SetupWhat the setup figure measured

There is no index. Preparation is segmentation only — 55 statements cut into 385 sections by src/segment.py, pure code, no model and no key.

LicenceLicence

MIT — this repository's own licence. Every account holder, institution and deposit is invented; a real bank or brokerage statement cannot be published at all — it is one of the most sensitive documents a person holds.

Bring your ownBring your own statements

Replace data/corpus/*.txt, write data/fields.json, and supply a gold record per statement with a stated flag for every optional deposit field. SECTION_HINTS in src/select.py maps fields to headings and will need editing for a different statement layout; when it does not match, selection falls back to the whole document — slower, more expensive, always correct.

Step 05 of 14Prompt

Written for: engineer · Logged from the run, verbatim.

In brief

The assembled prompt in full, and its real decomposition. Every part below occurs in the prompt that was actually sent, and the token counts sum to the recorded input total — they were measured against the provider’s own counter, not apportioned by character share.

AssemblyHow the prompt is assembled

This is the cost lesson as arithmetic: of the 863 tokens assembled, 360 are instructions — 42% of the prompt, the largest single part. A per-request bill follows whichever part is biggest, and on this kit that is the one named here.

Measured token-for-token via evals/prompt_tokens.py's nested-prefix subtraction against the provider's own tokenizer, not estimated from characters — see results/tokens-p001-asset-verify.json.

Full promptThe prompt in full

SYSTEM:
You extract structured fields from a bank or brokerage statement. You return JSON and nothing else.

RULES, in order of importance:
1. If the statement does not state a field, return null for it. Do not infer it, do not compute it, and do not use what you know about the world. A null is a correct answer -- a statement with no deposits this period correctly has null for every deposit field.
2. `deposit_documented` means the largest deposit's description names a SPECIFIC, VERIFIABLE INSTITUTIONAL counterparty: a named employer, "IRS TREAS", a named pension fund, "SSA"/Social Security, or a named bank/brokerage on a wire or transfer. It is 'no' for a personal P2P transfer (Zelle, Venmo, cash app), a cash deposit, a mobile check deposit with no payer named, or a generic "Deposit"/"Payroll Deposit"/"Direct Deposit" line that does NOT name an employer -- the word 'payroll' alone is not a counterparty. Read the whole description; do not decide from one keyword.
3. Copy values verbatim from the statement wherever possible.
4. Use the exact allowed value for a field that lists them.
5. Return every field named in the schema, even when the answer is null.

USER:
Extract these fields:
- account_holder (string) -- the name on the statement, verbatim
- institution (string) -- the bank or brokerage that issued the statement
- account_type (enum) one of: checking, savings, money_market, brokerage -- the account type stated on the statement
- period_start (string) -- the statement period's first date, YYYY-MM-DD
- period_end (string) -- the statement period's last date, YYYY-MM-DD
- beginning_balance (number) -- the stated beginning balance, no currency symbol
- ending_balance (number) -- the stated ending balance, no currency symbol
- largest_deposit_amount (number) -- the dollar amount of the single largest deposit line this period, or null if no deposits posted
- largest_deposit_description (string) -- that deposit's description, copied verbatim, or null if no deposits posted
- deposit_documented (enum) one of: yes, no -- does that deposit's description name a specific, verifiable INSTITUTIONAL counterparty (a named employer, IRS, a named pension fund, SSA, or a named bank/brokerage on a transfer)? A personal P2P transfer, cash deposit, or generic description is 'no' even if it uses the word payroll or direct deposit with no employer named. null if no deposits posted.

Return a JSON object with exactly these keys: account_holder, institution, account_type, period_start, period_end, beginning_balance, ending_balance, largest_deposit_amount, largest_deposit_description, deposit_documented
Use null for any field the statement does not state.

STATEMENT
---------
Statement Holder
----------------
Amara W. Haddad

Institution
-----------
Beacon Point Bank

Account Type
------------
savings

Statement Period
----------------
2026-06-01 to 2026-06-30

Account Summary
---------------
Beginning Balance: $18941.17
Ending Balance: $18099.20

Deposits This Period
--------------------
2026-06-07   $133.03   Mobile Check Deposit
2026-06-10   $168.01   ATM Deposit
2026-06-15   $238.78   Zelle from Amara N.
2026-06-19   $185.39   Mobile Check Deposit

Raw responseThe raw response

{
 "account_holder": "Amara W. Haddad",
 "institution": "Beacon Point Bank",
 "account_type": "savings",
 "period_start": "2026-06-01",
 "period_end": "2026-06-30",
 "beginning_balance": 18941.17,
 "ending_balance": 18099.2,
 "largest_deposit_amount": 238.78,
 "largest_deposit_description": "Zelle from Amara N.",
 "deposit_documented": "no"
}

Step 06 of 14Evals

Written for: the skeptic · The harness output, with as_of / model version / dataset version / verified_by.

In brief

Check a loan applicant's bank statements for large deposits — 55 statements. Two tiers of one model family answered, and every answer was then graded Two different ways — because the grade is a measurement too, and most of what went wrong on this kit went wrong in the ruler rather than the model.

MeasurementsWhat was measured

Two words carry this page: Counted is deterministic and nobody's opinion and Not yet known is printed blank rather than filled with something plausible. There is no Judged row here and that is the point: nothing on this page is a model's opinion about another model. The row most worth having is currently the empty one.

Grading costWhat it costs

Every dollar on these pages is a MEASURED token count multiplied by a named vendor's PUBLISHED rate -- a projection onto a rate card, not a bill anyone paid.

Rates checked 2026-08-18. The provider that actually ran r001 and r002 publishes no rate card this repo commits, so nothing here is what was actually paid -- the real spend for this kit's build is recorded in the commit history and the shared call ledger, not on this page.

The gradersTwo ways to grade

Every row opens its own page: how the test was run, the input on one real row, the prompt or formula with every iteration of it, and the analysis.

Choosing oneWhich one to use

The comparison is only useful if it ends in a choice. The answer is conditional — and it is not the grader with the best agreement score.

This pipeline was also attacked, which is a different question from whether its answers are right: see Threat model for the injection experiment, its two gates and what it does not prove.

Failure causesHow it fails

Step 07 of 14Unit cost

Written for: whoever signs off on the bill · Measured over the eval run, per model. Never quoted from a vendor page.

In brief

The token counts are measured on a real run and belong to this pipeline — which sections are selected, the field schema, the prompt — so they hold wherever you run it. The dollars are those counts multiplied by each vendor's published rate, checked 2026-08-18. Nobody paid these particular bills; they are what the same work would cost you on a card you can sign up to today.

Identical work in both rows — the same query, the same tokens. Only the price list changed. Which is why the token count is the measurement and the dollar is arithmetic you can redo on your own card.

Answering vs gradingAnswering and grading are two separate bills

The keyword floor (evals/baseline.py) and the scorer (evals/judge.py) are both pure code and cost $0.00 to run against either result set. The figure above is both runs' own token counts (the fast-tier and deliberating-tier runs, 55 statements each) priced at Google Gemini 3 Flash's published rate -- the same basis cost_per_query_usd uses, not a second, larger spend.

Cost driversWhat actually moves the bill

• The fixed system prompt and field schema outweigh the statement itself: 650 of 863 tokens (75%) on the example call. The floor every call pays before a single deposit line is even read.
• Output length: the model returns a full ten-key JSON record every call regardless of how many fields the statement actually states, so a statement with no deposits still pays for three null values in the reply.

Your volumeWhat it costs at your volume

Linear in statements: each call is independent and self-contained, with no shared context or index to amortise. This run's 55 statements cost about $0.072 projected onto Gemini 3 Flash's rate, so ten times the set is about $0.72 on the same rate and the same prompt — arithmetic on the measured per-call rate, not a second run.

Other modelsOn other models

This is arithmetic, not a run. No model below was executed against the labelled set, and no accuracy is claimed for any of them. What transfers is the input volume: retrieval is model-independent, so every model would receive the same passages.

The exact work behind every number on these pages: 55 bank or brokerage statements extracted, each scored by pure code against a mechanically-derived gold set. This run (r001-asset-verify, the fast tier) answered all 55 of 55 statements with no truncation -- the row this table prices, the same one Cost.cost_by_model[0] uses.

The codeEvery file, and what it is for

Five modules, no framework, and almost no dependencies. The pipeline is stdlib Python, because a kit is meant to be read and changed by someone who cloned it ten minutes ago — every dependency is a thing they have to understand before they can touch the part they came for.

Four of these are swap seams: the points you edit to make the kit yours. Each is marked, and each block below is the module’s real shape, read out of the file rather than written here.

The files on this page are not in the public GitHub catalog.

src/segment.pysegment

cut the statement into addressable sections, pure code

# Cut a document into addressable sections. Pure code — no model, no network.

def sections(text):
def locate(text, value):
def span_label(secs, start):

src/select.pyselect — a swap seam

pick which sections carry each field, pure code

You change it to: map fields to your own statement's headings; unmatched falls back to the whole document

# Pick which sections plausibly carry each field. Pure code -- the last deterministic step

SECTION_HINTS = {
def for_field(secs, field):
def plan(secs, fields):

src/prompt.pyprompt

assemble one call for all ten fields

# Assemble the extraction prompt. One prompt per statement, all ten fields in it.

SYSTEM = (
def field_schema(fields):
def build(doc_text, secs, fields, selector):
def parse(raw, fields):

src/extract.pyextract — a swap seam

the AI layer, one provider one key — plus the two pure-code computations downstream: reserve value and the large-deposit flag

You change it to: this kit's own flat threshold and vesting percentages — replace with your actual loan program's published rules before trusting the computed figures for anything real

# Extract one statement's fields: segment, select, prompt, one model call, then two pure-code

HERE = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
FIELDS = os.path.join(HERE, "data", "fields.json")
CORPUS = os.path.join(HERE, "data", "corpus")
MAX_TOKENS = 4000
LARGE_DEPOSIT_THRESHOLD_USD = 1000.0
VESTING = {"checking": 1.0, "savings": 1.0, "money_market": 1.0, "brokerage": 0.70}
def load_fields():
def load_doc(stmt_id):
def documents():

evals/judge.pyjudge

score field accuracy and the flag's recall/precision separately, pure code

# Score an extraction run. PURE CODE -- gold is exact and the answer is one value per cell, so

def norm(v):
def _num(v):
def equal(field, got, want):
def score_cell(field, got, want, stated):
def score(fields, records, golds):
def score_flags(flags, golds):

Start hereThe shortest path into it

Every entry above is a file in this kit, listed in the order the pipeline runs it.

ScenariosWhat it costs at your volume

Everything below is computed from one measured base: 840 input and 295 output tokens per query, all of it fixed — this kit sends the document's own selected sections, so there is no retrieval depth to turn. Change the volume and the bill scales linearly, because nothing here batches.

OperationsFour shapes of workload

The same arithmetic, at volumes worth naming. These are illustrations of scale, not measurements of anyone’s deployment.

Model follows the controls above.

What this ignoresWhere a real bill diverges

In one lineWhat this kit exposes

This run's statements are entirely synthetic (tools/build_corpus.py) -- no untrusted party wrote any statement text. In a real deployment a statement arrives from a bank, brokerage or the applicant's own upload -- exactly the kind of externally-authored input this kit's architecture reads verbatim and trusts, with no verification step before the deposit descriptions reach the model. No attack has been tried against this kit; see redteam.why.

Read from the shared .env or the real environment only, never written into the repo, never requested from a reader on any surface. src/app.py's /api/extract handler strips api_key and base_url out of any exception message before it reaches the browser, so a misconfigured key cannot leak into a UI error.

The experimentWe did not attack it — and three of four boundaries hold

An indirect prompt injection needs a field an outside party controls that reaches the prompt. In THIS corpus every statement is generated by tools/build_corpus.py, so there is no live untrusted text in the run this report measures -- but a real deployment's statements arrive from a bank, brokerage or the applicant's own upload, and that surface has not been attacked. The four gates below are boundaries confirmed by reading the code, not payloads run through it. Confirmed by reading the code, not by a run, on 2026-08-20 -- no attack run was fired; see redteam.why.

The first three boundaries hold, confirmed by reading the code, not by an attack trial. The fourth is the one this run has not tested: deposit_documented is the field the flag depends on entirely (see Guardrails.is_not), and no crafted input has ever been run against it.

The result0 attack trials, and three of four boundaries checked here hold in code: no write path exists, the large-deposit threshold and vesting percentages are fixed constants the model cannot move, and a misconfigured key cannot leak into a UI error. The fourth -- whether a crafted deposit description could talk the model into a fabricated deposit_documented=yes -- is unmeasured.

This run's corpus is entirely generated (tools/build_corpus.py) -- no statement's text was authored by an outside party. A real deployment's statements arrive from a bank, brokerage or the applicant's own upload -- exactly the kind of externally-supplied text this kit's architecture reads verbatim and trusts, with no verification step. Whether a deposit description crafted to mimic an institutional counterparty (a fake 'IRS TREAS' or named-employer string with no real basis) could talk the model into a false deposit_documented=yes is unmeasured for this kit.

HonestyWhat this does not prove

• Whether a deposit description crafted to mimic an institutional counterparty (a fake 'IRS TREAS' or named-employer string with no real basis in the statement) could talk the model into a false deposit_documented=yes -- no red-team run exists for this kit.
• Whether the live app's own /api/extract behaves identically to the registered run under an adversarial statement -- both use the same src/extract.py::extract(), but neither has been tested against one.
• Whether a code-level consistency check on deposit_documented (re-deriving it from the statement text by pure code) would catch a model's fabricated 'yes' in practice -- none has been built or exercised.

In one lineThe guardrail is enforced in code

Not a prompt instruction the model can ignore — a comparison applied after the reply is parsed, before anything downstream sees it.

A statement's largest deposit is flagged for underwriter review when it is $1,000 or more and deposit_documented is 'no'. Computed in pure code from the model's own extracted fields -- never asked of the model directly, never overridden by anything it returns.

src/extract.py::compute() -- LARGE_DEPOSIT_THRESHOLD_USD and the flag arithmetic are module-level constants and pure code, read once at import time. Unlike a prompt instruction the model could ignore, this guardrail cannot be talked out of firing by anything in the reply text; it only depends on the model's own deposit_documented value being right in the first place -- a DIFFERENT, prompt-level dependency the flag inherits (see is_not and could_not_verify).

EvidenceDoes it hold?

The limitWhat a guardrail is not

• IT IS NOT A CHECK ON deposit_documented ITSELF. The flag trusts the model's own answer for that field completely -- nothing re-derives deposit_documented from the statement text before computing the flag. The one repeatable disagreement this run has (self-paid interest, see Business.not_good_enough) is exactly the kind of error this guardrail cannot catch, because it never happened to cross the $1,000 threshold on this corpus -- not because the guardrail checked and cleared it.
• It does not guarantee an approval or a denial -- the flag is a routing signal for a human reviewer, never itself a decision. See UI.shots and Data.breaks_on.
• It has not been attacked. Whether a description crafted to move deposit_documented without crossing an obvious keyword could also evade the flag is unmeasured -- see Security.could_not_verify.

WatchedWhat is watched, and why that one

The kit measures 14 watched figures plus 8 comparability guards and 5 reference constants — three duties, not one list. 0 of the watched are grader rates, each on its own grader page; this board watches the ones that decide when to look again. Each square is one watched figure; filled means the latest run measured it.

8 measured by the latest run 6 need the model half

GuardsBefore any comparison

Preconditions, not alarms. If one moves, the runs were not measuring the same system and nothing else on this board may be differenced.

all shared guards held between the recorded runs (documents 55, extraction_cells 523, failures 0, refusal_cells 27, thinking True) — which is why the history below is allowed to mean anything.

The reference constants — the null-grader baseline, the adjudication matrix, the negative count — are the third duty. They are never re-measured per run: a baseline that moves with the thing it measures is not a baseline.

BandsThe alarm bands

Derived from evidence, never picked: exact-match where two runs reproduced a metric to the digit, wider than the measured same-input drift for latency, one row of the negative class for any rate — denominator printed beside it.

HistoryRun history

2 recorded runs. A dash is a metric that run did not measure, and a dash is never differenced: not measured is a third state, neither a value nor a zero, and reading it as zero is how a board reports an improvement on a run that never looked.

not a time series No two of these 2 runs measured the same system — they differ on the model — so nothing here may be differenced and the Verdict column is left off. Read this block DOWN a column, as a comparison between those runs, not along a row as a history.

DeviationsWhat deviated

0 breaches across 2 runs

The day a band breaks, a row renders here carrying the metric, the two runs, the band it crossed and what changed between them — model, prompt, chunker or top_k, read off the run records’ guards. A deviation with no diff beside it is an alarm nobody can act on.

RippleThe ripple map

Which figures move together, so one alarm reads as a system. Every edge carries its basis: measured names the run that showed it; reasoning admits nothing has varied it yet.

FiresWhat fires when a band breaks

Where evals stop sitting beside the guardrail and start feeding it. Where the answer is nothing automatic, the row says so — hiding it would undo “What a guardrail is not” above.

Nothing fires yet. Not one of the 6 bands on this kit names a consequence, so there is no row to print. Which is the honest answer to the heading: a band that nothing is wired to is a number somebody has to notice, and this kit has only those. An empty table would have said the same thing while looking like a rendering failure.

NextThe three you would add first

None of these is built here. They are named against the seam they sit on so they can be argued with.

CadenceWhen to re-run, and what it costs

Re-check the flag's evidence on any change to LARGE_DEPOSIT_THRESHOLD_USD or VESTING in src/extract.py, or to the deposit_documented rule in src/prompt.py -- either changes what the flag is computed FROM. Re-run evals/run.py (paid) on any change to MAX_TOKENS or the provider/model.

In one lineNo framework, deliberately

No framework and no dependencies beyond the standard library -- requirements.txt is empty, with a comment explaining that the emptiness is load-bearing. The whole extraction decision is three files: src/segment.py, src/select.py and src/adapters/__init__.py.

The mappingSeam by seam, who owns what

The seams are the ones this kit already publishes, so the mapping cannot drift from the code.

GraphsWhere LangGraph changes the answer

Nothing here is a graph. The pipeline has one path per statement -- segment, select, prompt, call, parse, compute -- with no branching and no state carried between statements. A framework would add an orchestrator to a for-loop.

The other sideWhat a framework costs you

• A different field schema (data/fields.json) or a different documented/undocumented rule needs its own gold set and its own evals/check_labels.py pass -- a framework's own schema layer would not remove that work, only relocate it.
• SECTION_HINTS is a five-minute edit for a new statement layout because it is a plain dict, not a configured retriever -- a framework's own chunking/retrieval abstraction would need its own re-tuning pass instead, with its own failure modes to learn.
• Swapping providers is one function and one entry in PROVIDERS (src/adapters/__init__.py) -- a framework's own model abstraction would add a dependency and a version to track for the same one-line change this file already gives away free.

In one lineWhat one run of this kit actually cost, in time and tokens

Measured on r001-asset-verify on the fast tier, 2026-08-20. This kit records telemetry measured per run — 4 of the 6 readings on this axis, 4 of them with a band the guardrail board can judge against.

The readingWhat the last run measured

No movement column. Not one of the 1 earlier run on record is comparable with this one — a different model tier, or a changed guard — and differencing across that measures two systems rather than one. Said once here rather than as a dash on every row.

Retrieval not shown. No run of this kit has ever recorded a retrieval timing, because the pipeline has no such stage. Those readings are left out rather than dashed: a dash says “this run did not measure it”, which promises a later run could.

DriftMedian call, run over run

Hatched bars are runs that are not comparable with the latest — a different model tier or a changed guard. They are drawn because hiding them would make the history look shorter than it is, and greyed because differencing across them would be measuring two systems.

The other sideWhat this does not tell you

The first of those is the one worth doing something about, and it is a solved problem: instrumenting a pipeline so each step reports its own duration is a couple of lines. Observability / tracing has the code, and what the result looks like once it is running.

In one lineWhat you provision, honestly

Everything measured on this page says where it was measured; everything not measured says that — the same contract as the other thirteen pages. Where a retrieval index would be on other kits, this one has: no index — each document goes whole into one call.

The machineWhat this needs

Enumerated from the kit’s own source by build/facts/envscan.py, committed, and drift-checked — never typed into this page.

Your dataWhere everything sits

The wireWhat crosses it, and when

Zero-egress topology: point BASE_URL at a local OpenAI-compatible server and nothing leaves the machine — statically true of the shipped adapter, whose only credential travels in the Authorization header.

The shapeHow it ships, and what you add

The ladderEach decision, and what you provision past its ceiling

Every threshold is a measurement with its provenance beside it, or says is_measured:false. A ceiling with no seam says so — it is the point the kit is outgrown, never a forced join.

When it failsFailure signatures

The knobsThe config surface, enumerated

Environment variables: API_KEY, BASE_URL, EMBED_MODEL, MAX_CALLS_PER_DAY, MODEL, PORT, PROVIDER.

The other sideWhat this page does not claim

The corpus licence, from the Data lens: MIT — this repository's own licence. Every account holder, institution and deposit is invented; a real bank or brokerage statement cannot be published at all — it is one of the most sensitive documents a person holds. Your corpus’s licence is yours to verify, and the labels you write are about your documents.

In one linePer-field exact match, light normalisation

Does the model's value for each field match gold, after trimming whitespace/punctuation and treating numbers within half a cent as equal?

MethodHow the test was run

evals/judge.py::score, in-process, no key and no model. The same function evals/baseline.py and evals/run.py both call -- a baseline and a real run scored by two different scorers cannot be compared honestly.

Every grader on these pages scored the same 110 already-recorded answers. Nothing was re-generated, so this compares rulers and not models.

The inputOne real row, seen by every grader

The formulaWhat it computes

The analysisWhat it actually did

In operationWhat to monitor

Reference standard: tools/build_corpus.py's gold, read back off the same generated deposit lines the document states (never a target that seeded them); evals/check_labels.py asserts stated/value consistency on every row before any run is allowed to spend.

How tight can the band be? There is no tolerance band on the field grade itself -- it is exact match after trimming whitespace/punctuation and treating numbers within half a cent as equal, never a continuous score to round.

Cadence: Re-score on any change to src/prompt.py, data/fields.json or tools/build_corpus.py -- the first two change what is asked, the third changes what is asked about. Re-run evals/run.py (paid) on any change to MAX_TOKENS or the provider/model.

The decisionWhen to reach for it

In one lineLarge-deposit flag confusion matrix

Does the run's own pure-code large_deposit_flag (computed from what the model extracted) match gold's large_deposit_flag?

MethodHow the test was run

evals/judge.py::score_flag, in-process, no key and no model. Computed from the run's own pure-code large_deposit_flag (src/extract.py::compute), never from a second model call -- the model never sees or states the flag directly.

Every grader on these pages scored the same 110 already-recorded answers. Nothing was re-generated, so this compares rulers and not models.

The inputOne real row, seen by every grader

The formulaWhat it computes

The analysisWhat it actually did

In operationWhat to monitor

Reference standard: tools/build_corpus.py's gold large_deposit_flag, computed the identical way src/extract.py::compute() computes it -- from the same generated deposit lines the document states, using LARGE_DEPOSIT_THRESHOLD_USD=1000.0.

How tight can the band be? There is no tolerance band -- the flag is a binary confusion matrix (flagged/not flagged against gold), never a continuous quantity to round.

Cadence: Re-score on any change to LARGE_DEPOSIT_THRESHOLD_USD in src/extract.py or tools/build_corpus.py, or to data/fields.json's deposit fields. Re-run evals/run.py (paid) on any change to MAX_TOKENS or the provider/model.

The decisionWhen to reach for it