cecli/cecli/models.py at 9f3dc156b27869cc525b10e0c2b22a545662f0bb · cecli-dev/cecli · GitHub

Repository navigation

History

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

322

323

324

325

326

327

328

329

330

331

332

333

334

335

336

337

338

339

340

341

342

343

344

345

346

347

348

349

350

351

352

353

354

355

356

357

358

359

360

361

362

363

364

365

366

367

368

369

370

371

372

373

374

375

376

377

378

379

380

381

382

383

384

385

386

387

388

389

390

391

392

393

394

395

396

397

398

399

400

401

402

403

404

405

406

407

408

409

410

411

412

413

414

415

416

417

418

419

420

421

422

423

424

425

426

427

428

429

430

431

432

433

434

435

436

437

438

439

440

441

442

443

444

445

446

447

448

449

450

451

452

453

454

455

456

457

458

459

460

461

462

463

464

465

466

467

468

469

470

471

472

473

474

475

476

477

478

479

480

481

482

483

484

485

486

487

488

489

490

491

492

493

494

495

496

497

498

499

500

501

502

503

504

505

506

507

508

509

510

511

512

513

514

515

516

517

518

519

520

521

522

523

524

525

526

527

528

529

530

531

532

533

534

535

536

537

538

539

540

541

542

543

544

545

546

547

548

549

550

551

552

553

554

555

556

557

558

559

560

561

562

563

564

565

566

567

568

569

570

571

572

573

574

575

576

577

578

579

580

581

582

583

584

585

586

587

588

589

590

591

592

593

594

595

596

597

598

599

600

601

602

603

604

605

606

607

608

609

610

611

612

613

614

615

616

617

618

619

620

621

622

623

624

625

626

627

628

629

630

631

632

633

634

635

636

637

638

639

640

641

642

643

644

645

646

647

648

649

650

651

652

653

654

655

656

657

658

659

660

661

662

663

664

665

666

667

668

669

670

671

672

673

674

675

676

677

678

679

680

681

682

683

684

685

686

687

688

689

690

691

692

693

694

695

696

697

698

699

700

701

702

703

704

705

706

707

708

709

710

711

712

713

714

715

716

717

718

719

720

721

722

723

724

725

726

727

728

729

730

731

732

733

734

735

736

737

738

739

740

741

742

743

744

745

746

747

748

749

750

751

752

753

754

755

756

757

758

759

760

761

762

763

764

765

766

767

768

769

770

771

772

773

774

775

776

777

778

779

780

781

782

783

784

785

786

787

788

789

790

791

792

793

794

795

796

797

798

799

800

801

802

803

804

805

806

807

808

809

810

811

812

813

814

815

816

817

818

819

820

821

822

823

824

825

826

827

828

829

830

831

832

833

834

835

836

837

838

839

840

841

842

843

844

845

846

847

848

849

850

851

852

853

854

855

856

857

858

859

860

861

862

863

864

865

866

867

868

869

870

871

872

873

874

875

876

877

878

879

880

881

882

883

884

885

886

887

888

889

890

891

892

893

894

895

896

897

898

899

900

901

902

903

904

905

906

907

908

909

910

911

912

913

914

915

916

917

918

919

920

921

922

923

924

925

926

927

928

929

930

931

932

933

934

935

936

937

938

939

940

941

942

943

944

945

946

947

948

949

950

951

952

953

954

955

956

957

958

959

960

961

962

963

964

965

966

967

968

969

970

971

972

973

974

975

976

977

978

979

980

981

982

983

984

985

986

987

988

989

990

991

992

993

994

995

996

997

998

999

1000

import asyncio

import importlib.resources

import json

import os

import platform

import sys

import time

from dataclasses import dataclass, fields

from pathlib import Path

from typing import Optional, Union

from uuid import uuid4 as generate_unique_id

import yaml

from cecli import __version__, utils

from cecli.decoding import safe_open

from cecli.dump import dump

from cecli.exceptions import LiteLLMExceptions

from cecli.helpers import coroutines, nested

from cecli.helpers.file_searcher import (

generate_search_path_list,

handle_core_files,

)

from cecli.helpers.model_config import get_default_config

from cecli.helpers.model_config.registry import register_default

from cecli.helpers.model_config.utils import get_entry_from_raw

from cecli.helpers.model_providers import ModelProviderManager

from cecli.helpers.nested import deep_merge

from cecli.helpers.requests import model_request_parser

from cecli.llm import litellm

from cecli.sendchat import sanity_check_messages

from cecli.utils import check_pip_install_extra

GLOBAL_ID = str(generate_unique_id())

RETRY_TIMEOUT = 60

COPY_PASTE_PREFIX = "cp:"

request_timeout = 600

DEFAULT_MODEL_NAME = "gpt-4o"

ANTHROPIC_BETA_HEADER = "prompt-caching-2024-07-31,pdfs-2024-09-25"

OPENAI_MODELS = """

o1-preview

o1-mini

o3-mini

gpt-4

gpt-4o

gpt-4o-2024-05-13

gpt-4-turbo-preview

gpt-4-0314

gpt-4-0613

gpt-4-32k

gpt-4-32k-0314

gpt-4-32k-0613

gpt-4-turbo

gpt-4-turbo-2024-04-09

gpt-4-1106-preview

gpt-4-0125-preview

gpt-4-vision-preview

gpt-4-1106-vision-preview

gpt-4o-mini

gpt-4o-mini-2024-07-18

gpt-3.5-turbo

gpt-3.5-turbo-0301

gpt-3.5-turbo-0613

gpt-3.5-turbo-1106

gpt-3.5-turbo-0125

gpt-3.5-turbo-16k

gpt-3.5-turbo-16k-0613

"""

OPENAI_MODELS = [ln.strip() for ln in OPENAI_MODELS.splitlines() if ln.strip()]

ANTHROPIC_MODELS = """

claude-2

claude-2.1

claude-3-haiku-20240307

claude-3-5-haiku-20241022

claude-3-opus-20240229

claude-3-sonnet-20240229

claude-3-5-sonnet-20240620

claude-3-5-sonnet-20241022

claude-sonnet-4-20250514

claude-opus-4-20250514

"""

ANTHROPIC_MODELS = [ln.strip() for ln in ANTHROPIC_MODELS.splitlines() if ln.strip()]

MODEL_ALIASES = {

"sonnet": "anthropic/claude-sonnet-4-20250514",

"haiku": "claude-3-5-haiku-20241022",

"opus": "claude-opus-4-20250514",

"4": "gpt-4-0613",

"4o": "gpt-4o",

"4-turbo": "gpt-4-1106-preview",

"35turbo": "gpt-3.5-turbo",

"35-turbo": "gpt-3.5-turbo",

"3": "gpt-3.5-turbo",

"deepseek": "deepseek/deepseek-chat",

"flash": "gemini/gemini-2.5-flash",

"flash-lite": "gemini/gemini-2.5-flash-lite",

"quasar": "openrouter/openrouter/quasar-alpha",

"r1": "deepseek/deepseek-reasoner",

"gemini-2.5-pro": "gemini/gemini-2.5-pro",

"gemini-3-pro-preview": "gemini/gemini-3-pro-preview",

"gemini": "gemini/gemini-3-pro-preview",

"gemini-exp": "gemini/gemini-2.5-pro-exp-03-25",

"grok3": "xai/grok-3-beta",

"optimus": "openrouter/openrouter/optimus-alpha",

}

@dataclass

class ModelSettings:

name: str

edit_format: str = "diff"

weak_model_name: Optional[str] = None

agent_model_name: Optional[str] = None

use_repo_map: bool = False

send_undo_reply: bool = False

lazy: bool = False

overeager: bool = False

reminder: str = "user"

examples_as_sys_msg: bool = False

extra_params: Optional[dict] = None

cache_control: bool = False

caches_by_default: bool = False

uses_messages_api: bool = False

use_system_prompt: bool = True

use_temperature: Union[bool, float] = True

streaming: bool = True

editor_model_name: Optional[str] = None

editor_edit_format: Optional[str] = None

reasoning_tag: Optional[str] = None

remove_reasoning: Optional[str] = None

system_prompt_prefix: Optional[str] = None

accepts_settings: Optional[list] = None

retries: Optional[dict] = None

retry_backoff_factor: float = 1.5

retry_on_unavailable: bool = True

retry_timeout: float = 30

request_timeout: int = request_timeout

debug: bool = False

MODEL_SETTINGS = []

with importlib.resources.open_text("cecli.resources", "model-settings.yml") as f:

model_settings_list = yaml.safe_load(f)

for model_settings_dict in model_settings_list:

MODEL_SETTINGS.append(ModelSettings(**model_settings_dict))

class ModelInfoManager:

CACHE_TTL = 60 * 60 * 24

def __init__(self):

self.cache_dir = handle_core_files(Path.home() / ".cecli" / "caches")

self.cache_file = self.cache_dir / "model_prices_and_context_window.json"

self.content = None

self._raw_content = None

self.local_model_metadata = {}

self.metadata_files = []

self.verify_ssl = True

self._cache_loaded = False

self.provider_manager = ModelProviderManager()

self.openai_provider_manager = self.provider_manager

def set_verify_ssl(self, verify_ssl):

self.verify_ssl = verify_ssl

self.provider_manager.set_verify_ssl(verify_ssl)

def _load_cache(self):

if self._cache_loaded:

return

try:

self.cache_dir.mkdir(parents=True, exist_ok=True)

if self.cache_file.exists():

cache_age = time.time() - self.cache_file.stat().st_mtime

if cache_age < self.CACHE_TTL:

try:

self._raw_content = self.cache_file.read_text()

except json.JSONDecodeError:

self._raw_content = None

except OSError:

pass

self._cache_loaded = True

def _get_entry_from_raw(self, key):

"""Parse a single model entry from raw JSON string without loading the entire dict."""

return get_entry_from_raw(self._raw_content, key)

def get_model_from_cached_json_db(self, model):

data = self.local_model_metadata.get(model)

if data:

return data

self._load_cache()

if not self._raw_content:

return dict()

info = self._get_entry_from_raw(model)

if info:

return info

pieces = model.split("/")

if len(pieces) == 2:

info = self._get_entry_from_raw(pieces[1])

if info and info.get("litellm_provider") == pieces[0]:

return info

return dict()

def get_raw_metadata(self):

"""Return the cached raw model metadata JSON string, or ``None``.

The model config pipeline consumes this raw string directly so the

metadata file is only ever held in memory once (no duplicate storage).

"""

self._load_cache()

return self._raw_content

def get_metadata_sources(self):

"""Return the metadata sources the model config pipeline should scan.

Prefers the model metadata files loaded via ``register_litellm_models``

(later files win, so user-supplied files override the bundled

resource), then the cached raw metadata string, else ``None`` so the

pipeline falls back to its bundled resource.

"""

if self.metadata_files:

return list(self.metadata_files)

self._load_cache()

return self._raw_content

def get_model_info(self, model):

cached_info = self.get_model_from_cached_json_db(model)

if cached_info:

return cached_info

litellm_info = None

if litellm._lazy_module or not cached_info:

try:

litellm_info = litellm.get_model_info(model)

except Exception as ex:

if "model_prices_and_context_window.json" not in str(ex):

print(str(ex))

provider_info = self._resolve_via_provider(model, cached_info)

if provider_info:

return provider_info

if litellm_info:

return litellm_info

if not cached_info and model.startswith("openai/"):

stripped = model[7:]

if stripped:

return self.get_model_info(stripped)

return cached_info

def _resolve_via_provider(self, model, cached_info):

if cached_info:

return None

provider = model.split("/", 1)[0] if "/" in model else None

if not self.provider_manager.supports_provider(provider):

return None

provider_info = self.provider_manager.get_model_info(model)

if provider_info:

self._record_dynamic_model(model, provider_info)

return provider_info

if provider == "openrouter":

openrouter_info = self.fetch_openrouter_model_info(model)

if openrouter_info:

openrouter_info.setdefault("litellm_provider", "openrouter")

self._record_dynamic_model(model, openrouter_info)

return openrouter_info

return None

def _record_dynamic_model(self, model, info):

self.local_model_metadata[model] = info

self._ensure_model_settings_entry(model)

def _ensure_model_settings_entry(self, model):

if any(ms.name == model for ms in MODEL_SETTINGS):

return

MODEL_SETTINGS.append(ModelSettings(name=model))

def fetch_openrouter_model_info(self, model):

"""

Fetch model info by scraping the openrouter model page.

Expected URL: https://openrouter.ai/<model_route>

Example: openrouter/qwen/qwen-2.5-72b-instruct:free

Returns a dict with keys: max_tokens, max_input_tokens, max_output_tokens,

input_cost_per_token, output_cost_per_token.

"""

url_part = model[len("openrouter/") :]

url = "https://openrouter.ai/" + url_part

try:

import requests

response = requests.get(url, timeout=5, verify=self.verify_ssl)

if response.status_code != 200:

return {}

html = response.text

import re

if re.search(

f"The model\\s*.*{re.escape(url_part)}.* is not available",

html,

re.IGNORECASE,

print(f"\x1b[91mError: Model '{url_part}' is not available\x1b[0m")

return {}

text = re.sub("<[^>]+>", " ", html)

context_match = re.search("([\\d,]+)\\s*context", text)

if context_match:

context_str = context_match.group(1).replace(",", "")

context_size = int(context_str)

else:

context_size = None

input_cost_match = re.search("\\$\\s*([\\d.]+)\\s*/M input tokens", text, re.IGNORECASE)

output_cost_match = re.search(

"\\$\\s*([\\d.]+)\\s*/M output tokens", text, re.IGNORECASE

)

input_cost = float(input_cost_match.group(1)) / 1000000 if input_cost_match else None

output_cost = float(output_cost_match.group(1)) / 1000000 if output_cost_match else None

if context_size is None or input_cost is None or output_cost is None:

return {}

params = {

"max_input_tokens": context_size,

"max_tokens": context_size,

"max_output_tokens": context_size,

"input_cost_per_token": input_cost,

"output_cost_per_token": output_cost,

"litellm_provider": "openrouter",

}

return params

except Exception as e:

print("Error fetching openrouter info:", str(e))

return {}

model_info_manager = ModelInfoManager()

class ModelOverrides:

"""Class-level manager for model tag overrides (suffix-based model name resolution).

Allows users to define model tag overrides like "model:suffix" that map to

a base model name with specific configuration overrides.

State is stored at the class level so it can be loaded once (e.g. from a YAML file

in the entry point) and automatically applied when Model is instantiated.

"""

_overrides: dict = {}

_defaults_overrides: dict = {}

@classmethod

def load_from_file(cls, git_root, model_overrides_fname, io, verbose=False):

"""Load model tag overrides from a YAML file."""

from pathlib import Path

import yaml

model_overrides_files = generate_search_path_list(

".cecli.model.overrides.yml", git_root, model_overrides_fname

)

overrides = {}

files_loaded = []

for fname in model_overrides_files:

try:

if Path(fname).exists():

with safe_open(fname, "r") as f:

content = yaml.safe_load(f)

if content:

for model_name, tags in content.items():

if model_name not in overrides:

overrides[model_name] = {}

overrides[model_name].update(tags)

files_loaded.append(fname)

except Exception as e:

io.tool_error(f"Error loading model overrides from {fname}: {e}")

if len(files_loaded) > 0 and verbose:

io.tool_output("Loaded model overrides from:")

for file_loaded in files_loaded:

io.tool_output(f" - {file_loaded}")

if (

model_overrides_fname

and model_overrides_fname not in files_loaded

and model_overrides_fname != ".cecli.model.overrides.yml"

io.tool_warning(f"Model Overrides File Not Found: {model_overrides_fname}")

cls._from_file_or_string(overrides)

@classmethod

def load_from_string(cls, model_overrides_str, io):

"""Load model tag overrides from a JSON/YAML string."""

import json

import yaml

overrides = {}

if not model_overrides_str:

return

try:

content = json.loads(model_overrides_str)

except json.JSONDecodeError:

content = yaml.safe_load(model_overrides_str)

if content and isinstance(content, dict):

for model_name, tags in content.items():

if model_name not in overrides:

overrides[model_name] = {}

overrides[model_name].update(tags)

except Exception as e:

io.tool_error(f"Error parsing model overrides string: {e}")

return

cls._from_file_or_string(overrides)

@classmethod

def _from_file_or_string(cls, overrides):

"""Internal: process raw overrides dict into class-level state."""

# Extract 'defaults' key for direct model name matching

defaults_overrides = {}

if "defaults" in overrides:

defaults_overrides = overrides.pop("defaults")

if not isinstance(defaults_overrides, dict):

defaults_overrides = {}

# Merge into class-level state

for model_name, tags in overrides.items():

if model_name not in cls._overrides:

cls._overrides[model_name] = {}

if isinstance(tags, dict):

cls._overrides[model_name].update(tags)

# Merge defaults

for model_name, tags in defaults_overrides.items():

if model_name not in cls._defaults_overrides:

cls._defaults_overrides[model_name] = {}

if isinstance(tags, dict):

cls._defaults_overrides[model_name].update(tags)

@classmethod

def apply(cls, model_name):

"""Return (effective_model_name, override_kwargs) for a given model_name.

If model_name ends with ":suffix" where suffix is configured for the

prefix (everything before the last colon), we return the prefix model

and the associated override dict. Otherwise we leave the name unchanged

and return empty overrides.

NOTE: COPY_PASTE_PREFIX handling is done by Model.__init__ before this

method is called, so we only work with the clean model name here.

"""

if not model_name:

return model_name, {}

# Try to find a matching override by checking all possible suffix matches.

# We iterate from right to left splitting on colons to handle cases where

# the base model name itself contains colons (e.g. "provider/model:tag:alias")

parts = model_name.split(":")

for i in range(len(parts) - 1, 0, -1):

potential_base = ":".join(parts[:i])

potential_suffix = ":".join(parts[i:])

# Check if this base has the suffix configured

if potential_base in cls._overrides:

suffixes = cls._overrides[potential_base]

if isinstance(suffixes, dict) and potential_suffix in suffixes:

cfg = suffixes[potential_suffix]

if isinstance(cfg, dict):

return potential_base, cfg.copy()

# Check for direct match in defaults overrides

if model_name in cls._defaults_overrides:

cfg = cls._defaults_overrides[model_name]

if isinstance(cfg, dict):

return model_name, cfg.copy()

# No match found

return model_name, {}

@classmethod

def clear(cls):

"""Reset all model overrides."""

cls._overrides = {}

cls._defaults_overrides = {}

class Model(ModelSettings):

def __init__(

self,

model,

from_model=None,

**kwargs,

provided_model = model or ""

if isinstance(provided_model, Model):

provided_model = provided_model.name

elif not isinstance(provided_model, str):

provided_model = str(provided_model)

io = kwargs.get("io", nested.getter(from_model, "io", None))

verbose = kwargs.get("verbose", nested.getter(from_model, "verbose", False))

override_kwargs = kwargs.get("override_kwargs", None)

retries = kwargs.get("retries", nested.getter(from_model, "retries", None))

debug = kwargs.get("debug", nested.getter(from_model, "debug", False))

if kwargs.get("sub_models", True):

agent_model = kwargs.get("agent_model", nested.getter(from_model, "agent_model", None))

weak_model = kwargs.get("weak_model", nested.getter(from_model, "weak_model", None))

editor_model = kwargs.get(

"editor_model", nested.getter(from_model, "editor_model", None)

)

editor_edit_format = kwargs.get(

"editor_edit_format",

nested.getter(from_model, "editor_edit_format", None),

)

else:

agent_model = kwargs.get("agent_model", None)

weak_model = kwargs.get("weak_model", None)

editor_model = kwargs.get("editor_model", None)

editor_edit_format = kwargs.get("editor_edit_format", None)

self.io = io

self.verbose = verbose

self.override_kwargs = override_kwargs or {}

self._default_reasoning_effort = None

self._default_thinking_budget = None

self.copy_paste_mode = False

self.copy_paste_transport = "api"

if provided_model.startswith(COPY_PASTE_PREFIX):

model = provided_model.removeprefix(COPY_PASTE_PREFIX)

self.enable_copy_paste_mode(transport="clipboard")

else:

model = provided_model

model = MODEL_ALIASES.get(model, model)

# Resolve model tag overrides (suffix-based model name resolution)

resolved_model, resolved_overrides = ModelOverrides.apply(model)

if resolved_overrides:

model = resolved_model

merged = resolved_overrides.copy()

if override_kwargs:

merged.update(override_kwargs)

override_kwargs = merged

self.override_kwargs = override_kwargs

self.name = model

self.max_chat_history_tokens = 1024

self.weak_model = None

self.editor_model = None

self.agent_model = None

self.extra_model_settings = next(

(ms for ms in MODEL_SETTINGS if ms.name == "cecli/extra_params"),

None,

)

self.info = self.get_model_info(model)

self.model_config_defaults = get_default_config(

model, model_info_manager.get_metadata_sources()

)

# Fill any gaps in the model info from the metadata-derived defaults;

# values already known (e.g. from litellm) win.

self.info = {**self.model_config_defaults.get("llm", {}), **self.info}

self.litellm_provider = (self.info.get("litellm_provider") or "").lower()

res = self.validate_environment()

self.missing_keys = res.get("missing_keys")

self.keys_in_environment = res.get("keys_in_environment")

max_input_tokens = self.info.get("max_input_tokens") or 0

self.max_chat_history_tokens = min(max(max_input_tokens / 16, 1024), 8192)

self.configure_model_settings(model)

self._apply_provider_defaults()

self._apply_reasoning_defaults()

self.get_weak_model(weak_model)

self.get_agent_model(agent_model)

self.retries = retries

self.debug = debug

if editor_model is False:

self.editor_model_name = None

else:

self.get_editor_model(editor_model, editor_edit_format)

if self.copy_paste_transport == "clipboard":

self.streaming = False

def get_model_info(self, model):

return model_info_manager.get_model_info(model)

def _copy_fields(self, source):

"""Helper to copy fields from a ModelSettings instance to self"""

for field in fields(ModelSettings):

val = getattr(source, field.name)

setattr(self, field.name, val)

if self.reasoning_tag is None and self.remove_reasoning is not None:

self.reasoning_tag = self.remove_reasoning

def configure_model_settings(self, model):

exact_match = False

for ms in MODEL_SETTINGS:

if model == ms.name:

self._copy_fields(ms)

exact_match = True

break

if self.accepts_settings is None:

self.accepts_settings = []

model = model.lower()

if not exact_match:

self.apply_generic_model_settings(model)

if (

self.extra_model_settings

and self.extra_model_settings.extra_params

and self.extra_model_settings.name == "cecli/extra_params"

if not self.extra_params:

self.extra_params = {}

for key, value in self.extra_model_settings.extra_params.items():

if isinstance(value, dict) and isinstance(self.extra_params.get(key), dict):

self.extra_params[key] = {**self.extra_params[key], **value}

else:

self.extra_params[key] = value

if self.name.startswith("openrouter/"):

if self.accepts_settings is None:

self.accepts_settings = []

if "thinking_tokens" not in self.accepts_settings:

self.accepts_settings.append("thinking_tokens")

if "reasoning_effort" not in self.accepts_settings:

self.accepts_settings.append("reasoning_effort")

# Apply metadata-derived defaults from the model config pipeline before

# user-supplied override kwargs so explicit configuration wins. The llm

# block was already merged into ``self.info`` (existing values win), so

# only the api/agent defaults are applied here.

pipeline_defaults = dict(self.model_config_defaults)

# helpers carries callables (e.g. format_reasoning), not extra_params.

pipeline_defaults.pop("helpers", None)

self._apply_structured_kwargs(pipeline_defaults, self.name)

if self.override_kwargs:

self._apply_structured_kwargs(self.override_kwargs, model)

def _apply_structured_kwargs(self, config, model_name):

"""Apply api/llm/agent override groups to info, extra_params and settings.

Used for both the metadata-derived defaults (model config pipeline) and

user-supplied override kwargs so both share the same application rules.

"""

if not config:

return

if not self.extra_params:

self.extra_params = {}

valid_model_settings_fields = {f.name for f in fields(ModelSettings)}

# Detect structured keys: api_settings, api, llm_settings, or llm keys

# indicate the new format.

has_structured_keys = any(

k in config

for k in (

"api_settings",

"api-settings",

"llm_settings",

"llm-settings",

"api",

"llm",

"agent",

)

for key, value in config.items():

if key in ("agent", "model_settings", "model-settings"):

if not isinstance(value, dict):

raise ValueError(

f"override_kwargs '{key}' must be a dict, got" f" {type(value)}"

)

for setting_key, setting_value in value.items():

if setting_key not in valid_model_settings_fields:

raise ValueError(

f"Invalid model_settings key '{setting_key}'. Must"

f" be one of: {sorted(valid_model_settings_fields)}"

)

setattr(self, setting_key, setting_value)

elif has_structured_keys and key in (

"api",

"api_settings",

"api-settings",

# api_settings: merge each sub-key into extra_params

if not isinstance(value, dict):

raise ValueError(f"override_kwargs '{key}' must be a dict, got {type(value)}")

for api_key, api_value in value.items():

if api_key == "reasoning_effort":

# Applied via set_reasoning_effort at init; remember the

# default so later (user-supplied) values win.

self._default_reasoning_effort = api_value

continue

if api_key == "thinking":

if isinstance(api_value, dict):

self._default_thinking_budget = api_value.get("budget_tokens")

else:

self._default_thinking_budget = api_value

continue

if isinstance(api_value, dict) and isinstance(

self.extra_params.get(api_key), dict

self.extra_params[api_key] = {

**self.extra_params[api_key],

**api_value,

}

else:

self.extra_params[api_key] = api_value

elif has_structured_keys and key in (

"llm",

"llm_settings",

"llm-settings",

# llm_settings: merge into self.info

if not isinstance(value, dict):

raise ValueError(f"override_kwargs '{key}' must be a dict, got {type(value)}")

self.info = {**self.info, **value}

if getattr(litellm, "model_cost", None) is not None:

if not litellm.model_cost.get(model_name):

litellm.model_cost[model_name] = {}

litellm.model_cost[model_name].update(self.info)

elif isinstance(value, dict) and isinstance(self.extra_params.get(key), dict):

self.extra_params[key] = {**self.extra_params[key], **value}

else:

self.extra_params[key] = value

def apply_generic_model_settings(self, model):

if "/o3-mini" in model:

self.edit_format = "diff"

self.use_repo_map = True

self.use_temperature = False

self.system_prompt_prefix = "Formatting re-enabled. "

if "reasoning_effort" not in self.accepts_settings:

self.accepts_settings.append("reasoning_effort")

return

if "gpt-4.1-mini" in model:

self.edit_format = "diff"

self.use_repo_map = True

self.reminder = "sys"

self.examples_as_sys_msg = False

return

if "gpt-4.1" in model:

self.edit_format = "diff"

self.use_repo_map = True

self.reminder = "sys"

self.examples_as_sys_msg = False

return

last_segment = model.split("/")[-1]

if last_segment in ("gpt-5", "gpt-5-2025-08-07") or "gpt-5.1" in model:

self.use_temperature = False

self.edit_format = "diff"

if "reasoning_effort" not in self.accepts_settings:

self.accepts_settings.append("reasoning_effort")

return

if "/o1-mini" in model:

self.use_repo_map = True

self.use_temperature = False

self.use_system_prompt = False

return

if "/o1-preview" in model:

self.edit_format = "diff"

self.use_repo_map = True

self.use_temperature = False

self.use_system_prompt = False

return

if "/o1" in model:

self.edit_format = "diff"

self.use_repo_map = True

self.use_temperature = False

self.streaming = False

self.system_prompt_prefix = "Formatting re-enabled. "

if "reasoning_effort" not in self.accepts_settings:

self.accepts_settings.append("reasoning_effort")

return

if "deepseek" in model and "v3" in model:

self.edit_format = "diff"

self.use_repo_map = True

self.reminder = "sys"

self.examples_as_sys_msg = True

return

if "deepseek" in model and ("r1" in model or "reasoning" in model):

self.edit_format = "diff"

self.use_repo_map = True

self.examples_as_sys_msg = True

self.use_temperature = False

self.reasoning_tag = "think"

return

if ("llama3" in model or "llama-3" in model) and "70b" in model:

self.edit_format = "diff"

self.use_repo_map = True

self.send_undo_reply = True

self.examples_as_sys_msg = True

return

if "gpt-4-turbo" in model or "gpt-4-" in model and "-preview" in model:

self.edit_format = "udiff"

self.use_repo_map = True

self.send_undo_reply = True

return

if "gpt-4" in model or "claude-3-opus" in model:

self.edit_format = "diff"

self.use_repo_map = True

self.send_undo_reply = True

return

if "gpt-3.5" in model or "gpt-4" in model:

self.reminder = "sys"

return

if "3-7-sonnet" in model:

self.edit_format = "diff"

self.use_repo_map = True

self.examples_as_sys_msg = True

self.reminder = "user"

if "thinking_tokens" not in self.accepts_settings:

self.accepts_settings.append("thinking_tokens")

return

if "3.5-sonnet" in model or "3-5-sonnet" in model:

self.edit_format = "diff"

self.use_repo_map = True

self.examples_as_sys_msg = True

self.reminder = "user"

return

if model.startswith("o1-") or "/o1-" in model:

self.use_system_prompt = False

self.use_temperature = False

return

if (

"qwen" in model

and "coder" in model

and ("2.5" in model or "2-5" in model)

and "32b" in model

self.edit_format = "diff"

self.editor_edit_format = "editor-diff"

self.use_repo_map = True

return

if "qwq" in model and "32b" in model and "preview" not in model:

self.edit_format = "diff"

self.editor_edit_format = "editor-diff"

self.use_repo_map = True

self.reasoning_tag = "think"

self.examples_as_sys_msg = True

self.use_temperature = 0.6

self.extra_params = dict(top_p=0.95)

return

if "qwen3" in model:

self.edit_format = "diff"

self.use_repo_map = True

if "235b" in model:

self.system_prompt_prefix = "/no_think"

self.use_temperature = 0.7

self.extra_params = {"top_p": 0.8, "top_k": 20, "min_p": 0.0}

else:

self.examples_as_sys_msg = True

self.use_temperature = 0.6

self.reasoning_tag = "think"

self.extra_params = {"top_p": 0.95, "top_k": 20, "min_p": 0.0}

return

if self.edit_format == "diff":

self.use_repo_map = True

return

def __str__(self):

return self.name

def enable_copy_paste_mode(self, *, transport="api"):

self.copy_paste_mode = True

self.copy_paste_transport = transport

def get_weak_model(self, provided_model):

if provided_model is False:

self.weak_model = self

self.weak_model_name = None

return

if self.copy_paste_transport == "clipboard":

self.weak_model = self

self.weak_model_name = None

return

if isinstance(provided_model, Model):

self.weak_model = provided_model

self.weak_model_name = provided_model.name

return

if provided_model:

self.weak_model_name = provided_model

if not self.weak_model_name:

self.weak_model = self

return

if self.weak_model_name == self.name:

self.weak_model = self

return

self.weak_model = Model(self.weak_model_name, from_model=self, sub_model=False)

return self.weak_model

def get_agent_model(self, provided_model):

if provided_model is False:

self.agent_model = self

self.agent_model_name = None

return

if self.copy_paste_transport == "clipboard":

self.agent_model = self

self.agent_model_name = None

return

if isinstance(provided_model, Model):

self.agent_model = provided_model

self.agent_model_name = provided_model.name

return

if provided_model:

self.agent_model_name = provided_model

if not self.agent_model_name:

self.agent_model = self

return

if self.agent_model_name == self.name:

self.agent_model = self

return

self.agent_model = Model(self.agent_model_name, from_model=self, sub_model=False)

return self.agent_model

def get_editor_model(self, provided_model, editor_edit_format):

if self.copy_paste_transport == "clipboard":

provided_model = False

self.editor_model_name = self.name

self.editor_model = self

if isinstance(provided_model, Model):

self.editor_model = provided_model

self.editor_model_name = provided_model.name

elif provided_model:

self.editor_model_name = provided_model

if editor_edit_format:

self.editor_edit_format = editor_edit_format

if not self.editor_model_name or self.editor_model_name == self.name:

self.editor_model = self

else:

self.editor_model = Model(self.editor_model_name, from_model=self, sub_model=False)

if not self.editor_edit_format:

self.editor_edit_format = self.editor_model.edit_format

if self.editor_edit_format in ("diff", "whole", "diff-fenced"):

self.editor_edit_format = "editor-" + self.editor_edit_format

return self.editor_model

def commit_message_models(self):

return [self.weak_model, self]

def _ensure_extra_params_dict(self):

if self.extra_params is None:

self.extra_params = {}

elif not isinstance(self.extra_params, dict):

self.extra_params = dict(self.extra_params)

def _apply_provider_defaults(self):

provider = self._configured_provider()

self.litellm_provider = provider or None

if self.info.get("supports_stream") is False:

self.streaming = False

if not provider:

return

provider_config = model_info_manager.provider_manager.get_provider_config(provider)

if not provider_config:

return

self._ensure_extra_params_dict()

self.extra_params.setdefault("custom_llm_provider", provider)

if provider_config.get("supports_stream") is False:

self.streaming = False

base_url = model_info_manager.provider_manager.get_provider_base_url(provider)

if base_url:

self.extra_params.setdefault("base_url", base_url)

default_headers = provider_config.get("default_headers") or {}

if default_headers:

headers = self.extra_params.setdefault("extra_headers", {})

for key, value in default_headers.items():

headers.setdefault(key, value)

provider_extra = provider_config.get("extra_params") or {}

for key, value in provider_extra.items():

if key not in self.extra_params:

self.extra_params[key] = value

def _apply_reasoning_defaults(self):

"""Apply the default thinking/reasoning configuration at init time.

The model config pipeline's ``api`` block carries the reasoning effort

and/or thinking budget the model uses; the config registry records the

defaults so they initialize from the api block or an explicit user

override. ``override_kwargs`` applied later in

``_apply_structured_kwargs`` win over these defaults.

"""

register_default(

self.name,

reasoning=self._default_reasoning_effort,

thinking=self._default_thinking_budget,

)

if self._default_reasoning_effort is not None:

self.set_reasoning_effort(self._default_reasoning_effort)

if self._default_thinking_budget is not None:

self.set_thinking_tokens(self._default_thinking_budget)

View remainder of file in raw view

cecli/cecli/models.py at 9f3dc156b27869cc525b10e0c2b22a545662f0bb · cecli-dev/cecli · GitHub

Repository navigation

Expand file tree

Breadcrumbs

models.py

Latest commit

History

Breadcrumbs

models.py

File metadata and controls