Image-Text-to-Text
MLX
Safetensors
qwen3_5
clairic
cloudflare
systemone
structured-output
information-extraction
multimodal
custom-code
conversational
4-bit precision
Instructions to use mlx-community/clairic-4bit with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- MLX
How to use mlx-community/clairic-4bit with MLX:
# Make sure mlx-vlm is installed # pip install --upgrade mlx-vlm from mlx_vlm import load, generate from mlx_vlm.prompt_utils import apply_chat_template from mlx_vlm.utils import load_config # Load the model model, processor = load("mlx-community/clairic-4bit") config = load_config("mlx-community/clairic-4bit") # Prepare input image = ["http://images.cocodataset.org/val2017/000000039769.jpg"] prompt = "Describe this image." # Apply chat template formatted_prompt = apply_chat_template( processor, config, prompt, num_images=1 ) # Generate output output = generate(model, processor, formatted_prompt, image) print(output) - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- LM Studio
- Pi
How to use mlx-community/clairic-4bit with Pi:
Start the MLX server
# Install MLX LM: uv tool install mlx-lm # Start a local OpenAI-compatible server: mlx_lm.server --model "mlx-community/clairic-4bit"
Configure the model in Pi
# Install Pi: npm install -g @earendil-works/pi-coding-agent # Add to ~/.pi/agent/models.json: { "providers": { "mlx-lm": { "baseUrl": "http://localhost:8080/v1", "api": "openai-completions", "apiKey": "none", "models": [ { "id": "mlx-community/clairic-4bit" } ] } } }Run Pi
# Start Pi in your project directory: pi
- Hermes Agent
How to use mlx-community/clairic-4bit with Hermes Agent:
Start the MLX server
# Install MLX LM: uv tool install mlx-lm # Start a local OpenAI-compatible server: mlx_lm.server --model "mlx-community/clairic-4bit"
Configure Hermes
# Install Hermes: curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash hermes setup # Point Hermes at the local server: hermes config set model.provider custom hermes config set model.base_url http://127.0.0.1:8080/v1 hermes config set model.default mlx-community/clairic-4bit
Run Hermes
hermes
- Atomic Chat
- OpenClaw
How to use mlx-community/clairic-4bit with OpenClaw:
Start the MLX server
# Install MLX LM: uv tool install mlx-lm # Start a local OpenAI-compatible server: mlx_lm.server --model "mlx-community/clairic-4bit"
Configure OpenClaw
# Install OpenClaw: npm install -g openclaw@latest # Register the local server and set it as the default model: openclaw onboard --non-interactive --mode local \ --auth-choice custom-api-key \ --custom-base-url http://127.0.0.1:8080/v1 \ --custom-model-id "mlx-community/clairic-4bit" \ --custom-provider-id mlx-lm \ --custom-compatibility openai \ --custom-text-input \ --accept-risk \ --skip-health
Run OpenClaw
openclaw agent --local --agent main --message "Hello from Hugging Face"
File size: 11,896 Bytes
1a48d6b | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 | from __future__ import annotations
import json
import re
from collections.abc import Iterable
from dataclasses import asdict, dataclass
from datetime import date
from typing import Any
VERSION = "source-date-candidates-v1"
INSTRUCTION = "Return the complete calendar date for this requested field as YYYY-MM-DD. Choose only a date explicitly printed for the requested role in the supplied source. Use the full day, month, and year, including legal ordinal date wording. Do not derive an unprinted date from a duration, term length, another date, or general knowledge. Return null if the requested role has no printed date. A date for another role does not fill this field."
_MONTHS = {
"jan": 1,
"january": 1,
"feb": 2,
"february": 2,
"mar": 3,
"march": 3,
"apr": 4,
"april": 4,
"may": 5,
"jun": 6,
"june": 6,
"jul": 7,
"july": 7,
"aug": 8,
"august": 8,
"sep": 9,
"sept": 9,
"september": 9,
"oct": 10,
"october": 10,
"nov": 11,
"november": 11,
"dec": 12,
"december": 12,
}
_MONTH = "(?:" + "|".join(sorted(_MONTHS, key=len, reverse=True)) + ")\\.?"
_DMY = re.compile(
"(?<!\\w)(?:the\\s+)?(?P<day>\\d{1,2})(?P<ordinal>st|nd|rd|th)?\\s+(?:(?:day\\s+)?of\\s+)?(?P<month>"
+ _MONTH
+ "),?\\s+(?P<year>\\d{4})(?!\\w)",
re.IGNORECASE,
)
_MDY = re.compile(
"(?<!\\w)(?P<month>"
+ _MONTH
+ ")\\s+(?:the\\s+)?(?P<day>\\d{1,2})(?P<ordinal>st|nd|rd|th)?,?\\s+(?P<year>\\d{4})(?!\\w)",
re.IGNORECASE,
)
_YMD = re.compile(
"(?<![\\w./-])(?P<year>\\d{4})(?P<sep>[-/.])(?P<month>\\d{1,2})(?P=sep)(?P<day>\\d{1,2})(?!(?:[\\w/-]|\\.\\w))"
)
_NUMERIC = re.compile(
"(?<![\\w./-])(?P<first>\\d{1,2})(?P<sep>[-/.])(?P<second>\\d{1,2})(?P=sep)(?P<year>\\d{4})(?!(?:[\\w/-]|\\.\\w))"
)
_DERIVED = re.compile(
"\\b(?:calculat\\w*|comput\\w*|deriv\\w*|infer\\w*|after|before|ago|next|previous|following|prior|add|subtract|plus|minus|earlier|later|offset|duration|anniversar\\w*|turns?)\\b|\\b(?:first|last)\\s+day\\s+of\\b|\\bnet\\s*\\d+\\b",
re.IGNORECASE,
)
_RELATIVE = re.compile(
"\\b(?:today|tomorrow|yesterday|now)\\b|\\bcurrent\\s+(?:calendar\\s+)?date\\b|\\bbusiness\\s+days?\\b|\\b(?:days?|weeks?|months?|years?)\\s+from\\b|\\b(?:this|last)\\s+(?:week|month|year)\\b|\\b(?:in|within)\\s+(?:\\d+|[a-z]+(?:[\\s-]+[a-z]+){0,2})\\s+(?:business\\s+)?(?:days?|weeks?|months?|years?)\\b",
re.IGNORECASE,
)
_READING_INTENT = re.compile(
"\\b(?:dates?|dated|calendar\\s+day|printed|explicit(?:ly)?|stated|shown|recorded|written|literal|verbatim)\\b",
re.IGNORECASE,
)
_EXPRESSION_REQUEST = re.compile(
"\\b(?:value\\s+(?:of|for)|evaluat\\w*|expression|formula|variable|constant|function|code)\\b",
re.IGNORECASE,
)
_CALENDAR_EXPRESSION = re.compile(
"\\b(?:add|sub(?:tract)?|plus|minus|offset|shift|set)[_]?(?:business[_]?)?(?:days?|weeks?|months?|years?)\\s*\\(|\\b(?:date_?(?:add|sub|diff)|timedelta|relativedelta|DateOffset|EDATE|EOMONTH|WORKDAY)\\s*\\(|\\.\\s*(?:set(?:UTC)?(?:Date|Month|FullYear)|add|subtract)\\s*\\(|\\binterval\\s+(?:['\\\"]|\\d)|\\b(?:date|datetime)(?:\\s*\\.\\s*\\w+)?\\s*\\([^;\\n]*\\)\\s*[+-]",
re.IGNORECASE,
)
def eligible(spec: dict[str, Any]) -> bool:
instructions = spec.get("instructions", "")
return (
spec.get("type") == "date"
and (not spec.get("knowledge"))
and ("source_only" not in spec)
and ("null_threshold" not in spec)
and isinstance(instructions, str)
and bool(_READING_INTENT.search(instructions))
and (
not (
_DERIVED.search(instructions)
or _RELATIVE.search(instructions)
or _EXPRESSION_REQUEST.search(instructions)
)
)
)
@dataclass(frozen=True)
class Candidate:
start: int
end: int
printed: str
iso_values: tuple[str, ...]
notation: str
@property
def ambiguous(self) -> bool:
return len(self.iso_values) != 1
def _iso(year: str | int, month: str | int, day: str | int) -> str | None:
try:
return date(int(year), int(month), int(day)).isoformat()
except ValueError:
return None
def _ordinal_valid(day: int, suffix: str | None) -> bool:
if not suffix:
return True
expected = (
"th"
if 11 <= day % 100 <= 13
else {1: "st", 2: "nd", 3: "rd"}.get(day % 10, "th")
)
return suffix.casefold() == expected
def candidates(
source: str, *, numeric_order: str | None = None
) -> tuple[Candidate, ...]:
if numeric_order not in (None, "mdy", "dmy"):
raise ValueError("Numeric date order must be explicitly mdy or dmy")
result = []
for pattern in (_DMY, _MDY):
for match in pattern.finditer(source):
if not _ordinal_valid(int(match["day"]), match["ordinal"]):
continue
value = _iso(
match["year"],
_MONTHS[match["month"].rstrip(".").casefold()],
match["day"],
)
if value:
result.append(
Candidate(
match.start(), match.end(), match[0], (value,), "named_month"
)
)
for match in _YMD.finditer(source):
value = _iso(match["year"], match["month"], match["day"])
if value:
result.append(
Candidate(match.start(), match.end(), match[0], (value,), "year_first")
)
for match in _NUMERIC.finditer(source):
(first, second) = (int(match["first"]), int(match["second"]))
readings = set()
orders = (numeric_order,) if numeric_order else ("mdy", "dmy")
for order in orders:
value = _iso(
match["year"],
first if order == "mdy" else second,
second if order == "mdy" else first,
)
if value:
readings.add(value)
if readings:
result.append(
Candidate(
match.start(),
match.end(),
match[0],
tuple(sorted(readings)),
"numeric",
)
)
return tuple(sorted(set(result), key=lambda item: (item.start, -item.end)))
class DateSource:
def __init__(self, source: str, *, numeric_order: str | None = None) -> None:
self.text = source
self.candidates = candidates(source, numeric_order=numeric_order)
self.ambiguous = tuple((item for item in self.candidates if item.ambiguous))
self.support: dict[str, list[Candidate]] = {}
for item in self.candidates:
if not item.ambiguous:
self.support.setdefault(item.iso_values[0], []).append(item)
self.values = tuple(self.support)
def constraint(self, *, nullable: bool = True) -> dict[str, str]:
literals = [json.dumps(value) for value in self.values]
if nullable:
literals.append("null")
if not literals:
raise ValueError(
"No unambiguous source date satisfies a nonnullable date contract"
)
return {
"ebnf": "root ::= " + " | ".join((json.dumps(value) for value in literals))
}
def span(self, value: object) -> tuple[bool, list[int] | None]:
matches = self.support.get(value, ()) if isinstance(value, str) else ()
intervals = {(item.start, item.end) for item in matches}
return (
bool(intervals),
list(next(iter(intervals))) if len(intervals) == 1 else None,
)
def evidence(self, value: str) -> list[dict[str, Any]]:
return [asdict(item) for item in self.support.get(value, ())]
def annotate_predictions(
self, predictions: dict[str, dict[str, Any]], names: Iterable[str]
) -> dict[str, dict[str, Any]]:
result = dict(predictions)
for name in names:
if name not in predictions:
continue
answer = dict(predictions[name])
result[name] = answer
answer.update(
date_decoder=VERSION, score_representation="native_json_iso_date_wire"
)
if (
not answer.get("valid")
or not answer.get("complete")
or answer.get("value") is None
):
continue
(found, span) = self.span(answer["value"])
if not found:
answer.update(
valid=False,
value=None,
span=None,
failure_reason="not_a_printed_date_candidate",
)
continue
answer.update(
grounding_method="printed_date_candidates_source",
span=span,
source_date_candidates=self.evidence(answer["value"]),
)
answer.pop("span_method", None)
if span is not None:
answer.update(
span_method="unique_normalized_printed_date_match",
printed_value=self.text[slice(*span)],
)
return result
def merge_predictions(
self,
original: dict[str, dict[str, Any]],
alternatives: dict[str, dict[str, Any]],
names: Iterable[str],
) -> dict[str, dict[str, Any]]:
result = original
if not self.automatic_route_safe():
return result
for name in names:
previous = original.get(name)
answer = alternatives.get(name)
if not isinstance(previous, dict) or not isinstance(answer, dict):
continue
if previous.get("type") not in (None, "date"):
continue
previous_value = previous.get("value")
if not isinstance(previous_value, str) or previous_value in self.support:
continue
value = answer.get("value")
if (
answer.get("valid") is not True
or answer.get("complete") is not True
or (not isinstance(value, str))
or (value not in self.support)
or (answer.get("date_decoder") != VERSION)
or (answer.get("grounding_method") != "printed_date_candidates_source")
or (answer.get("span") != self.span(value)[1])
or (not isinstance(answer.get("raw"), str))
):
continue
try:
if (
date.fromisoformat(value).isoformat() != value
or json.loads(answer["raw"]) != value
):
continue
except (ValueError, TypeError):
continue
if value == previous.get("value"):
continue
if result is original:
result = dict(original)
result[name] = {**answer, "type": "date"}
return result
def automatic_route_safe(self) -> bool:
return (
bool(self.values)
and (not self.ambiguous)
and (not _CALENDAR_EXPRESSION.search(self.text))
)
def allows(self, spec: dict[str, Any]) -> bool:
return eligible(spec) and self.automatic_route_safe()
def report(self) -> dict[str, Any]:
return {
"version": VERSION,
"values": self.values,
"candidates": [asdict(item) for item in self.candidates],
"ambiguous_candidates": [asdict(item) for item in self.ambiguous],
"calendar_expression_in_source": bool(
_CALENDAR_EXPRESSION.search(self.text)
),
"automatic_route_safe": self.automatic_route_safe(),
}
|