query_image
Answer free-form questions about any image using visual QA. Use it to read watermarks, logos, or stylized text, and get consistency checks that flag low-confidence or contradictory AI answers.
Instructions
Ask a free-form question about an image (visual question answering).
This is the right tool for reading photo watermarks, logos, signage,
or any cursive/stylized/low-contrast text — ask e.g. "What does the
text/watermark say, exactly?". The ocr tool misreads that kind of
text confidently; prefer this one for it instead.
Moondream2 is a small model and is documented to answer open-ended judgment
questions ("describe anything wrong in this image") with a flat "None" on
images that all had real visible defects, and to give the same yes/no answer
across genuinely different images -- a default response, not a real
observation. Set check_consistency=true to make that visible: the tool also
asks a rephrased control question and returns, per image,
{answer, control_answer, consistent, confidence}. confidence is "low"
in either of two failure modes: the two answers agree on a short
default-looking token ("None", "Yes", "Nothing", ...), which is the
signature of a flat default rather than a genuine observation; or they
substantively contradict each other, which makes either answer weaker
evidence than it looks alone. "normal" requires substantive answers that
agree. A low result on a judgment question means you should not trust the
answer without independent confirmation. When the answer is
low-confidence, the tool also tries to route to the measurement that
actually answers the question: it classifies the question's wording and,
if a measurable category applies and the object names parse from the
wording, runs that tool's measurement (spatial_relations for a
contact/containment question, count_objects for "how many", ocr for a
text-reading question, detect_objects for "which is largest/smallest")
and attaches it as cross_check. The "largest/smallest" case detects every
instance of the named object and picks the extremum by bounding-box area,
returning its box -- so "which circle is biggest" resolves to coordinates,
not a repeated guess. The cross-check is omitted when no measurement
applies or the names can't be parsed -- it never guesses.
Input Schema
| Name | Required | Description | Default |
|---|---|---|---|
| src | Yes | Local file path or http(s) URL of the image to process. PDFs are also accepted and are rendered one image per page, so tools that return a list return one entry per page. | |
| question | Yes | A free-form question to ask about the image. | |
| check_consistency | No | When true, also ask a rephrased control question and report whether the two answers agree. `confidence` is 'low' in two cases: the answers agree on a short default-looking token ('None', 'Yes', 'No', 'Nothing', ...), or they substantively contradict each other. Moondream2 is a small VLM that answers open-ended judgment questions ('describe anything wrong') with a flat 'None' on images that all had real visible defects -- this layer makes that default-answer behavior visible instead of presenting it as reliable. On a low-confidence answer it also routes to the measurement that actually answers the question when one applies (spatial_relations for a contact/containment question, count_objects for 'how many', ocr for a text-reading question) and attaches it as `cross_check`. Default false keeps the original list[str] return; true returns one {answer, control_answer, consistent, confidence, cross_check?} dict per image. | |
| structured_analysis | No | When true, bypasses the simple VQA return and returns structured Observation and Anomaly payloads per image. Corroborates VLM claims against physical Grounding DINO measurements to proactively detect anatomical and structural AI generation artifacts (e.g. extra/missing body parts). |
Output Schema
| Name | Required | Description | Default |
|---|---|---|---|
| result | Yes |