AI Demos Research — the structured-intelligence platform. Every verdict on these pages opens to the execution behind it.
Graded 12 November 2026

Can Dify retrieve wording from a plain digital text document?

Dify returned the passage that states the asked-for fact from a plain digital text document. Asked for the Tolvan A3's recommended recalibration interval in the document's own words, it ranked the whole Calibration passage first, which gives the interval as 22 months from the date of the last certificate.

1 of 1 test case passed

Every test case in this scenario has a result.

Pass rate100%1 of 1 gradable
Coverage1 of 1test case with a result
1PassDid everything it was expected to do.
0FailNo test case failed.
0Not gradableEvery result here could be graded.
0UntestedEvery test case in this scenario has a result.
The pass rate is a summary. The evidence is the test case below: what we sent, what we checked, what the tool returned, and the proof.

The test case

Each test case is judged on its own: Pass, Fail, Not gradable, or Untested. The scenario result above is based on this test case.

Plain prose fact, asked in the document's own wordsPassEvidence
What we sent

“What is the recommended recalibration interval for the Tolvan A3?”

InputFile (Markdown)
Input document: the Tolvan A3 product overview, with the interval in its Calibration section
Markdown55 lines
# Tolvan A3 - product overview

The Tolvan A3 is a conductivity and temperature sensor built for
continuous deployment in estuarine and coastal water. It is the instrument most
of our customers meet first, and most of the questions the service desk answers
are about this one.

## What it measures

The instrument reports conductivity, temperature and a derived practical
salinity. Conductivity is measured with an inductive cell, which has no exposed
electrodes to foul or to wear, and temperature with a thermistor mounted in the
same flow path so that the two readings describe the same parcel of water.
Pressure is not measured; where a depth record is needed, the instrument is
paired with a separate pressure sensor on the same mooring.

## Where it is used

Typical deployments are estuary monitoring for a regulator, intake monitoring
for a water treatment works, and baseline surveys ahead of construction in a
harbour. It is rated for continuous immersion to fifty metres. Deeper work uses
the Tolvan C5, which shares the sensing arrangement in a
pressure housing rated considerably deeper.

## How it is powered and read

On a mooring, the instrument runs from an internal battery pack and logs to
internal memory. On a cabled installation it is powered over the same cable
that carries its data, and reports continuously to the Brannoc
telemetry gateway, which forwards to the customer's own system. The two modes
are configured at commissioning and are not switched in the field.

## Fouling and cleaning

In estuarine water the practical limit on deployment is biological fouling
rather than power or memory. An anti-fouling guard and a sacrificial anode are
fitted at commissioning and replaced at each service visit. Customers in warm
or nutrient-rich water are advised to shorten the interval between visits
rather than to clean the sensing face themselves, because the inductive cell is
easily damaged by an abrasive.

## Calibration

Every instrument leaves the laboratory with a signed certificate recording the
as-found and as-left bias at four reference points across the working range.
The recommended recalibration interval for the Tolvan A3 is 22 months from the date of the last certificate. An instrument returned early because of a
suspected drift is checked as found before anything is adjusted, so that the
drift can be quantified rather than merely corrected.

## Firmware

Firmware is updated during a service visit rather than over the telemetry link,
because an interrupted update on a deployed instrument leaves it unreadable
until it is recovered. A customer who declines a recommended update is asked to
confirm in writing, and the confirmation is attached to the instrument record.
Open the Markdown file ↗
What the tool returned
Text
Calibration

Every instrument leaves the laboratory with a signed certificate recording the as-found and as-left bias at four reference points across the working range. The recommended recalibration interval for the Tolvan A3 is 22 months from the date of the last certificate. An instrument returned early because of a suspected drift is checked as found before anything is adjusted, so that the drift can be quantified rather than merely corrected.

Firmware

Retrieval log: the question sent and all four ranked passages returned

Copied from Output file (text), line 20

Open the text file ↗
Expected vs. Found
✓Expected: The passage that states the asked-for fact comes back.Found: The first of four results, scoring 0.769, was the whole Calibration passage stating a 22-month recalibration interval; the next-best result scored 0.579.
Supporting proof
Proof 1Screenshot
Notebook screenshot: the retrieval call and its printed ranked results
Open original ↗
Why this result

The test required the passage stating the fact to come back. The top result was the Calibration passage giving the 22-month interval, which the source document states only once, so the right passage was returned and the result is a Pass.

Also observed in this test case

Lower results held no competing interval. The three lower-ranked results were the product overview's introduction and two warranty documents stating 24- and 36-month warranty periods; none of them gave a recalibration interval, so no competing figure was returned.

Tested by Vivek Tomar · evidence dated 21 September 2026

Configuration and setup

How this tool was set up for the run and what the test needed in place. Each row is a fact from the run's records; a fact the records do not hold is left out, not guessed.

Software that produced the output
Dify
Build or version
Dify Cloud
Surface
API
Set up before the run
The document set had to be plain digital prose stating one fact exactly once: the Tolvan A3 product overview gives its recommended recalibration interval in its Calibration section. After ingestion, the product was asked for that interval in the document's own wording.
Tested
21 September 2026 · Vivek Tomar

How this scenario is graded

How we decide Pass, Fail and Not gradable. The same rules apply to every tool tested on this scenario.

How results are decided

Each test case gets one result per tool: Pass, Fail or Not gradable. A test case we haven't run yet shows Untested. There are no partial results.

The rules
  • Pass: the tool did everything the test expected, and nothing it said contradicts the correct answer.
  • Fail: at least one expected behaviour clearly didn't happen; the test case's result says which one and quotes the tool.
  • Not gradable: our evidence couldn't settle the outcome (for example a record we needed is missing). It is never counted as a fail, and the test case's result says what's missing.

Where this sits in the benchmark

This page is one cell of a larger study: one tool, one scenario. Only this benchmark's frame appears here.

LevelNameScope
BenchmarkManaged RAG →15 scenarios · 9 tools
CapabilityData Ingestion →
ScenarioAn ordinary digital text document →
ToolDify →

History of this result

What has happened to this result since it was first published. Runs and grades are never overwritten: a retest or a re-grade publishes a new result and keeps the earlier one readable.

from the publication record
16 November 2026First publishedManaged RAG v1

Act on this result

Nothing filed here edits the run or the grade. A challenge opens a review, and a review can produce a new run or a re-grade — which becomes the current result and leaves this one in the history.

This matches what I see

You run the same kind of test against your own setup and get the same behaviour.

Agree →
This does not match

Yours behaves differently. Tell us what you got, with a screenshot if you have one.

Disagree →
Point out an issue

Something here is wrong — a reference value, a transcription, a grade.

Report an issue →
Request a retest

On a newer build, a larger dataset, or your own setup.

Request a retest →
We have fixed this

Tell us what changed and we schedule a rerun of the failing test case. The old result stays as history.

Vendor notice →
Filed against this evidenceNothing yet. Challenges, counter-evidence and fix notices appear here with their outcome, and stay on the page after they are resolved.
Cite this result
aidemos.com/benchmarks/managed-rag/results/dify/an-ordinary-digital-text-document · 1 pass · 0 fail · coverage 1/1 · graded 2026-11-12

The same record is available as structured data through the AI Demos MCP server, with the counts, the coverage and every per-test-case reason carried as fields.

Verify the proof files

These files support this result. Open a file to inspect the original evidence.

Input · File (Markdown) ↗text/markdown; charset=utf-8 · 3 KB
Output file (text) ↗text/plain; charset=utf-8 · 7 KB
Proof 1 · Screenshot ↗image/png · 297 KB
File fingerprints (SHA-256)

A fingerprint identifies the exact file used for this result.

Input · File (Markdown)83fdf5e69b1bad5e271d7655313f48321c98f9f9a8961f63c44ad05656be1f84
Output file (text)e564367b6c8c3fea593fc8aa112b47afb0152082979d42d45e2d6335f1603330
Proof 1 · Screenshote1ba3055e4283cbae4001b32eef2c043fe40a3f210b3ef27756668b99981a3be