Image-Text-to-Text
Transformers
Safetensors
qwen3_5
clairic
cloudflare
systemone
multimodal
structured-output
information-extraction
custom-code
conversational
Instructions to use Cloudflare/clairic with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use Cloudflare/clairic with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("image-text-to-text", model="Cloudflare/clairic") messages = [ { "role": "user", "content": [ {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"}, {"type": "text", "text": "What animal is on the candy?"} ] }, ] pipe(text=messages)# pip install -U transformers accelerate # Load model directly from transformers import AutoProcessor, AutoModelForMultimodalLM processor = AutoProcessor.from_pretrained("Cloudflare/clairic") model = AutoModelForMultimodalLM.from_pretrained("Cloudflare/clairic", device_map="auto") messages = [ { "role": "user", "content": [ {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"}, {"type": "text", "text": "What animal is on the candy?"} ] }, ] inputs = processor.apply_chat_template( messages, add_generation_prompt=True, tokenize=True, return_dict=True, return_tensors="pt", ).to(model.device) outputs = model.generate(**inputs, max_new_tokens=256) print(processor.decode(outputs[0][inputs["input_ids"].shape[-1]:])) - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- vLLM
How to use Cloudflare/clairic with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "Cloudflare/clairic" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "Cloudflare/clairic", "messages": [ { "role": "user", "content": [ { "type": "text", "text": "Describe this image in one sentence." }, { "type": "image_url", "image_url": { "url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg" } } ] } ] }'Use Docker
docker model run hf.co/Cloudflare/clairic
- SGLang
How to use Cloudflare/clairic with SGLang:
Install from pip and serve model
# Install SGLang from pip: pip install sglang # Start the SGLang server: python3 -m sglang.launch_server \ --model-path "Cloudflare/clairic" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "Cloudflare/clairic", "messages": [ { "role": "user", "content": [ { "type": "text", "text": "Describe this image in one sentence." }, { "type": "image_url", "image_url": { "url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg" } } ] } ] }'Use Docker images
docker run --gpus all \ --shm-size 32g \ -p 30000:30000 \ -v ~/.cache/huggingface:/root/.cache/huggingface \ --env "HF_TOKEN=<secret>" \ --ipc=host \ lmsysorg/sglang:latest \ python3 -m sglang.launch_server \ --model-path "Cloudflare/clairic" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "Cloudflare/clairic", "messages": [ { "role": "user", "content": [ { "type": "text", "text": "Describe this image in one sentence." }, { "type": "image_url", "image_url": { "url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg" } } ] } ] }' - Docker Model Runner
How to use Cloudflare/clairic with Docker Model Runner:
docker model run hf.co/Cloudflare/clairic
Download clairic_runtime/printed_numbers.py from Cloudflare/clairic: direct link, hf CLI and curl.
- Browser
- Download file 10.1 kB
-
https://huggingface.co/Cloudflare/clairic/resolve/main/clairic_runtime/printed_numbers.py
- Command line
-
hf download hf://Cloudflare/clairic/clairic_runtime/printed_numbers.py
-
curl -L -o printed_numbers.py https://huggingface.co/Cloudflare/clairic/resolve/main/clairic_runtime/printed_numbers.py
10.1 kB
| from __future__ import annotations | |
| import math | |
| import re | |
| from collections.abc import Collection | |
| from decimal import Decimal, DecimalException | |
| from typing import Any | |
| from jsonschema import Draft202012Validator, ValidationError | |
| from . import ar | |
| COPY_INSTRUCTION = " Copy the complete requested number exactly as printed, including every digit, currency symbol, grouping separator, sign or parentheses. Do not calculate, normalize or round. Return null only if this requested value is absent." | |
| PAYMENT_PRESENCE_INSTRUCTION = " Every separately printed payment actually received is a present row, including an amount of zero, a negative amount, or a few cents. Copy all matching printed received-payment rows in their printed order, preserving duplicate rows. Do not return null or end the list while a matching received-payment row remains." | |
| _NUMBER = re.compile( | |
| "\\s*(?P<open>\\()?\\s*(?P<before>[+-]?)\\s*(?:[$€£¥]|USD|EUR|GBP|CAD|AUD|JPY)?\\s*(?P<after>[+-]?)\\s*(?P<number>(?:\\d{1,3}(?:,\\d{3})+|\\d+)(?:\\.\\d+)?(?:[eE][+-]?\\d+)?)\\s*(?P<close>\\))?\\s*", | |
| re.IGNORECASE, | |
| ) | |
| _TRANSFORM = re.compile( | |
| "\\b(?:calculat\\w*|comput\\w*|deriv\\w*|convert\\w*|summar\\w*|aggregat\\w*|group\\w*|combin\\w*|deduplicat\\w*|normaliz\\w*)\\b", | |
| re.IGNORECASE, | |
| ) | |
| def _literal(spec: dict[str, Any]) -> bool: | |
| instructions = spec.get("instructions", "") | |
| if not isinstance(instructions, str) or spec.get("knowledge"): | |
| return False | |
| instructions = re.sub( | |
| "\\b(?:do\\s+not|never|don't)\\s+(?:calculate\\w*|compute\\w*|derive\\w*|convert\\w*|summarize\\w*|aggregate\\w*|group\\w*|combine\\w*|deduplicate\\w*|normalize\\w*)\\b", | |
| "", | |
| instructions, | |
| flags=re.IGNORECASE, | |
| ) | |
| return not _TRANSFORM.search(instructions) | |
| def _payment_property(spec: dict[str, Any]) -> str | None: | |
| properties = spec.get("properties", {}) | |
| if ( | |
| spec.get("type") != "object" | |
| or len(properties) != 2 | |
| or (not _literal(spec)) | |
| or ( | |
| not all( | |
| ( | |
| isinstance(child, dict) and _literal(child) | |
| for child in properties.values() | |
| ) | |
| ) | |
| ) | |
| or (sum((child.get("type") == "text" for child in properties.values())) != 1) | |
| or (sum((child.get("type") == "number" for child in properties.values())) != 1) | |
| ): | |
| return None | |
| return next( | |
| (name for (name, child) in properties.items() if child.get("type") == "number") | |
| ) | |
| def payment_lists(fields: dict[str, dict[str, Any]]) -> list[str]: | |
| selected = [] | |
| for name, spec in fields.items(): | |
| item = spec.get("items") | |
| if ( | |
| spec.get("type") != "list" | |
| or not isinstance(item, dict) | |
| or (not _literal(spec)) | |
| or (_payment_property(item) is None) | |
| ): | |
| continue | |
| positive = re.split( | |
| "[.\\n]|\\bExclude\\b", | |
| spec.get("instructions", ""), | |
| maxsplit=1, | |
| flags=re.IGNORECASE, | |
| )[0] | |
| if ( | |
| re.search("\\bpayments?\\b[^.\\n]*\\breceived\\b", positive, re.IGNORECASE) | |
| or re.search( | |
| "\\breceived\\b[^.\\n]*\\bpayments?\\b", positive, re.IGNORECASE | |
| ) | |
| ) and ( | |
| not re.search( | |
| "\\b(?:not|never|exclude|excluding|ignore|without)\\b", | |
| positive, | |
| re.IGNORECASE, | |
| ) | |
| ): | |
| selected.append(name) | |
| return selected | |
| def normalize(value: object, spec: dict[str, Any]) -> int | float | None: | |
| if value is None: | |
| return None | |
| if not isinstance(value, str): | |
| raise TypeError("A printed number must be a string") | |
| raw = value.strip() | |
| unit = spec.get("unit") | |
| if unit and raw.endswith(unit): | |
| raw = raw[: -len(unit)].rstrip() | |
| match = _NUMBER.fullmatch(raw) | |
| if ( | |
| match is None | |
| or bool(match["open"]) != bool(match["close"]) | |
| or (match["before"] and match["after"]) | |
| or (match["open"] and (match["before"] or match["after"])) | |
| ): | |
| raise ValueError("Unsupported printed number") | |
| try: | |
| number = Decimal(match["number"].replace(",", "")) | |
| if not math.isfinite(float(number)): | |
| raise ValueError("Nonfinite numeric answer") | |
| except DecimalException as error: | |
| raise ValueError("Unsupported printed number") from error | |
| if match["open"] or "-" in (match["before"], match["after"]): | |
| number = number.copy_negate() | |
| result = ar.normalize_number(number, spec) | |
| if isinstance(result, float) and (not math.isfinite(result)): | |
| raise ValueError("Nonfinite numeric answer") | |
| Draft202012Validator(ar.schema(spec)).validate(result) | |
| return result | |
| def _text_spec(spec: dict[str, Any]) -> dict[str, Any]: | |
| return { | |
| "type": "text", | |
| "nullable": spec.get("nullable", True), | |
| "instructions": spec.get("instructions", "") + COPY_INSTRUCTION, | |
| } | |
| def prepare( | |
| fields: dict[str, dict[str, Any]], *, payment_rows: Collection[str] = () | |
| ) -> tuple[dict[str, dict[str, Any]], dict[str, dict[str, Any]]]: | |
| fields = {name: dict(spec) for (name, spec) in fields.items()} | |
| contracts = {} | |
| payment_rows = set(payment_rows) | |
| for parent in payment_lists(fields): | |
| prefix = parent + ".item" | |
| children = [ | |
| name | |
| for name in payment_rows | |
| if name.startswith(prefix) and name[len(prefix) :].isdigit() | |
| ] | |
| if not children: | |
| continue | |
| instructions = fields[parent].get("instructions", "") | |
| if PAYMENT_PRESENCE_INSTRUCTION in instructions: | |
| continue | |
| fields[parent] = { | |
| **fields[parent], | |
| "instructions": instructions + PAYMENT_PRESENCE_INSTRUCTION, | |
| } | |
| for child in children: | |
| if child not in fields or _payment_property(fields[child]) is None: | |
| continue | |
| requested = fields[child].get("instructions", "") | |
| fields[child] = { | |
| **fields[child], | |
| "instructions": requested.replace( | |
| instructions, instructions + PAYMENT_PRESENCE_INSTRUCTION, 1 | |
| ) | |
| if instructions and instructions in requested | |
| else requested + PAYMENT_PRESENCE_INSTRUCTION, | |
| } | |
| for name, spec in fields.items(): | |
| if spec["type"] == "number" and spec.get("source_only"): | |
| contracts[name] = {"scalar": spec} | |
| fields[name] = _text_spec(spec) | |
| elif ( | |
| spec["type"] == "list" | |
| and spec.get("source_only") | |
| and (spec["items"] == "number") | |
| ): | |
| fields[name] = { | |
| **spec, | |
| "items": "text", | |
| "instructions": spec.get("instructions", "") + COPY_INSTRUCTION, | |
| } | |
| elif ( | |
| name in payment_rows | |
| and (number_name := _payment_property(spec)) is not None | |
| ): | |
| number_spec = spec["properties"][number_name] | |
| contracts[name] = {"row": spec, "properties": {number_name: number_spec}} | |
| fields[name] = { | |
| **spec, | |
| "properties": { | |
| **spec["properties"], | |
| number_name: _text_spec(number_spec), | |
| }, | |
| } | |
| return (fields, contracts) | |
| def _normalize_row( | |
| raw: object, contract: dict[str, Any], record: dict[str, Any], name: str | |
| ) -> dict[str, Any] | None: | |
| if raw is None: | |
| result = None | |
| elif isinstance(raw, dict): | |
| result = dict(raw) | |
| current = record.get("fields", {}).get(name, {}).get("properties", {}) | |
| for property_name, spec in contract["properties"].items(): | |
| value = raw[property_name] | |
| if current.get(property_name, {}).get("type") == "number": | |
| if value is not None and ( | |
| isinstance(value, bool) | |
| or not isinstance(value, (int, float, Decimal)) | |
| ): | |
| raise TypeError("A numeric row value must be a number") | |
| result[property_name] = ar.normalize_number(value, spec) | |
| else: | |
| if value is not None and (not record.get("_media_payload")): | |
| if not isinstance(value, str) or value not in record["input"]: | |
| raise ValueError("not_in_source") | |
| result[property_name] = normalize(value, spec) | |
| else: | |
| raise TypeError("A printed payment row must be an object") | |
| Draft202012Validator(ar.schema(contract["row"])).validate(result) | |
| return result | |
| def normalize_predictions( | |
| record: dict[str, Any], predictions: dict[str, dict[str, Any]] | |
| ) -> dict[str, dict[str, Any]]: | |
| result = dict(predictions) | |
| for name, contract in record.get("_meta", {}).get("printed_numbers", {}).items(): | |
| answer = dict(result[name]) | |
| result[name] = answer | |
| if not answer.get("valid"): | |
| continue | |
| raw = answer["value"] | |
| try: | |
| if "properties" in contract: | |
| answer["value"] = _normalize_row(raw, contract, record, name) | |
| else: | |
| if raw is not None and (not record.get("_media_payload")): | |
| if not isinstance(raw, str) or raw not in record["input"]: | |
| raise ValueError("not_in_source") | |
| answer["value"] = normalize(raw, contract["scalar"]) | |
| answer["printed_value"] = raw | |
| answer["normalization_method"] = "printed_number_v1" | |
| except ( | |
| ValueError, | |
| TypeError, | |
| KeyError, | |
| OverflowError, | |
| ValidationError, | |
| ) as error: | |
| answer.update( | |
| value=None, | |
| valid=False, | |
| confidence=0.0, | |
| span=None, | |
| failure_reason="not_in_source" | |
| if str(error) == "not_in_source" | |
| else "invalid_printed_number", | |
| ) | |
| return result | |