Skip to content

艺术动画skill:35种艺术风格、9种解说语法,用代码让画动起来。

15.2 KB
Raw Blame
#!/usr/bin/env python3
"""图片接入:选择当前宿主工具、生成调用计划、验收并收纳素材。不代替Agent调用工具。"""
import argparse
import hashlib
import io
import json
import os
from pathlib import Path
import shutil
import sys
import time
import uuid

import capabilities as c


def sha(path):
    return hashlib.sha256(Path(path).read_bytes()).hexdigest()


def inspect_image(path):
    try:
        from PIL import Image
    except ImportError:
        raise c.ConfigError('图片验收需要Pillow;状态探查不需要。请按环境安装依赖') from None
    try:
        payload=Path(path).read_bytes()
        with Image.open(io.BytesIO(payload)) as im:
            fmt=im.format
            if fmt not in ('PNG','JPEG','WEBP'):
                raise c.ConfigError('支持PNG/JPEG/WEBP静态素材')
            if getattr(im,'n_frames',1)!=1:
                raise c.ConfigError('多帧图片请先明确选帧,不隐式截取')
            im.verify()
        with Image.open(io.BytesIO(payload)) as im:
            im.load()
            alpha=im.convert('RGBA').getchannel('A').getextrema()
            return {'width':im.width,'height':im.height,'format':fmt.lower(),
                    'has_transparency':alpha[0]<255,'has_visible_pixels':alpha[1]>0,'sha256':hashlib.sha256(payload).hexdigest()}
    except c.ConfigError:raise
    except Exception as e:
        raise c.ConfigError('图片无法完整解码:'+type(e).__name__) from None


def runtime_tools(path, session):
    data=c.read_json(Path(path))
    if data.get('session_id')!=session or not session:
        raise c.ConfigError('宿主能力快照不属于本次会话')
    if data.get('host') not in ('codex','claude-code','kimi','other'):
        raise c.ConfigError('未知宿主;可使用other但须提供真实工具声明')
    observed=data.get('observed_at')
    if not isinstance(observed,(int,float)) or not 0<=time.time()-observed<=900:
        raise c.ConfigError('工具快照已过期,请从当前会话重新取得(15分钟内)')
    tools=data.get('tools')
    if not isinstance(tools,list):raise c.ConfigError('工具快照需要tools列表;不可用时显式给空列表')
    ids=set()
    for tool in tools:
        if not isinstance(tool,dict) or not isinstance(tool.get('id'),str) or not tool['id'].strip() or tool['id'] in ids:
            raise c.ConfigError('工具ID缺失或重复')
        ids.add(tool['id'])
        if tool.get('kind') not in ('host','mcp') or tool.get('cost') not in ('subscription','paid_api','unknown','none') or tool.get('network') not in ('cloud','local'):
            raise c.ConfigError('工具必须声明执行来源、计费类型和网络范围')
        if not isinstance(tool.get('operations'),list) or not tool['operations'] or any(x not in ('generate','edit') for x in tool['operations']):
            raise c.ConfigError('工具操作声明无效')
        for key in ('reference_images','transparent_background'):
            if not isinstance(tool.get(key),bool):raise c.ConfigError('缺少工具特性:'+key)
    return data


def request_data(path):
    data=c.read_json(Path(path))
    if set(data)-{'prompt','operation','references','transparent_background','min_width','min_height'}:
        raise c.ConfigError('图片需求包含不支持字段')
    if not isinstance(data.get('prompt'),str) or not data['prompt'].strip():
        raise c.ConfigError('需要非空图片描述')
    data.setdefault('operation','generate');data.setdefault('references',[])
    data.setdefault('transparent_background',False);data.setdefault('min_width',1);data.setdefault('min_height',1)
    if data['operation'] not in ('generate','edit') or not isinstance(data['references'],list) or not isinstance(data['transparent_background'],bool):
        raise c.ConfigError('图片需求类型不正确')
    for key in ('min_width','min_height'):
        if type(data[key]) is not int or not 1<=data[key]<=16384:raise c.ConfigError('尺寸须为1—16384的整数')
    if data['operation']=='edit' and not data['references']:raise c.ConfigError('改图需要明确输入图片')
    refs=[]
    for item in data['references']:
        if not isinstance(item,str):raise c.ConfigError('参考图须为本地文件路径')
        p=Path(item).expanduser()
        if not p.is_absolute():p=Path(path).resolve().parent/p
        meta=inspect_image(p)
        refs.append({'path':str(p.resolve()),'sha256':meta['sha256']})
    data['references']=refs
    return data


def choose(config, runtime=None, request=None, mode=None, provider=None, grants=()):
    prefs=config['preferences']['image']; chosen=mode or prefs['mode']
    result={'capability':'image','mode':chosen,'status':'ready','reasons':[]}
    if config['policy']['image_generation']=='deny' and chosen=='generate':
        return dict(result,status='blocked',reasons=['图片生成已被明确禁止'])
    if chosen=='off':return dict(result,status='skipped',reasons=['不启用图片生成'])
    if chosen=='existing':return dict(result,status='needs_input',reasons=['提供现有图片,通过use-image导入'])
    if chosen=='unconfigured':return dict(result,status='needs_configuration',prompt=config['onboarding']['image']!='manual',reasons=['可选现有素材、当前宿主生图或跳过'])
    if chosen!='generate':raise c.ConfigError('图片方式不支持')
    if not runtime:return dict(result,status='needs_runtime',reasons=['需当前会话实际暴露的工具快照;不能按Agent名称推断'])
    request=request or {'operation':'generate','references':[],'transparent_background':False}
    selected=provider or prefs['provider']
    candidates=[]
    for tool in runtime['tools']:
        if selected!='auto' and tool['id']!=selected:continue
        if request['operation'] not in tool['operations']:continue
        if request['references'] and not tool['reference_images']:continue
        if request['transparent_background'] and not tool['transparent_background']:continue
        candidates.append(tool)
    if not candidates:return dict(result,status='unavailable',reasons=['当前无满足要求的工具;可导入现有素材,不自动接API'])
    ready=[];missing=[];blocked=[]
    for tool in candidates:
        needs=['image_generation']
        if tool['network']=='cloud':needs.append('cloud')
        if tool['cost']=='paid_api':needs.append('paid_api')
        if tool['cost']=='unknown':needs.append('unknown_image_cost')
        if request['references'] and tool['network']=='cloud':needs.append('reference_upload')
        deny=[k for k in needs if config['policy'][k]=='deny']
        ask=[k for k in needs if config['policy'][k]!='allow' and k not in grants]
        if deny:blocked.append({'tool_id':tool['id'],'denied':deny})
        elif ask:missing.append({'tool_id':tool['id'],'needs':ask})
        else:ready.append(tool)
    if len(ready)>1:return dict(result,status='needs_selection',candidates=[x['id'] for x in ready],reasons=['多个工具可用,请选择本次供应商或设置偏好'])
    if not ready:
        return dict(result,status='needs_permission' if missing else 'blocked',candidates=missing or blocked,reasons=['探查不授予调用权;沿用用户已经明确的本次授权'])
    return dict(result,status='host_action_required',host=runtime['host'],tool=ready[0],operation=request['operation'],
                reasons=['由当前Agent按真实工具schema执行;Python不会调用宿主工具'])


def no_overwrite(path, payload):
    path=Path(path);path.parent.mkdir(parents=True,exist_ok=True)
    with path.open('xb') as f:f.write(payload)


def config_fingerprint(config):
    selected={'preferences':config['preferences']['image'], 'policy':{k:config['policy'][k] for k in ('image_generation','cloud','paid_api','reference_upload','unknown_image_cost')}}
    return hashlib.sha256(json.dumps(selected,sort_keys=True).encode()).hexdigest()


def make_plan(config, request, runtime, session, out, mode=None, provider=None, grants=()):
    decision=choose(config,runtime,request,mode,provider,grants)
    if decision['status']!='host_action_required':return decision
    plan={'schema_version':1,'plan_id':str(uuid.uuid4()),'session_id':session,'created_at':time.time(),
          'host':runtime['host'],'tool':decision['tool'],'request':request,'configuration_sha256':config_fingerprint(config),
          'instruction':'Agent调用当前真实工具;成功后保存回执,accept-image重新核验授权与产物。'}
    no_overwrite(out,(json.dumps(plan,ensure_ascii=False,indent=2)+'\n').encode())
    return {'status':'host_action_required','plan':str(out),'plan_id':plan['plan_id'],'tool_id':plan['tool']['id']}


def accept(config, plan_path, receipt_path, file, out, runtime, session, grants=()):
    plan=c.read_json(Path(plan_path));receipt=c.read_json(Path(receipt_path))
    if plan.get('schema_version')!=1 or plan.get('session_id')!=session or plan.get('host')!=runtime['host']:
        raise c.ConfigError('计划不属于当前宿主/会话')
    created=plan.get('created_at')
    if not isinstance(created,(int,float)) or not 0<=time.time()-created<=3600:
        raise c.ConfigError('图片计划超过1小时,重新解析当前授权与要求')
    if plan.get('configuration_sha256')!=config_fingerprint(config):
        raise c.ConfigError('图片相关配置已改变,请重新生成计划')
    req=plan.get('request')
    if not isinstance(req,dict) or not isinstance(req.get('prompt'),str) or req.get('operation') not in ('generate','edit') or not isinstance(req.get('references'),list) or type(req.get('transparent_background')) is not bool:
        raise c.ConfigError('图片计划的需求结构无效')
    if any(type(req.get(k)) is not int or not 1<=req[k]<=16384 for k in ('min_width','min_height')):
        raise c.ConfigError('图片计划尺寸无效')
    for ref in req['references']:
        if not isinstance(ref,dict) or not isinstance(ref.get('path'),str) or not isinstance(ref.get('sha256'),str):
            raise c.ConfigError('计划参考图结构无效')
    tool=plan.get('tool',{})
    if not isinstance(tool,dict):raise c.ConfigError('计划工具声明无效')
    current=next((t for t in runtime['tools'] if t['id']==tool.get('id')),None)
    if current!=tool:raise c.ConfigError('工具能力发生变化或已不可用,不能复用旧计划')
    for ref in plan['request']['references']:
        if sha(ref['path'])!=ref['sha256']:raise c.ConfigError('参考图已变化,重新生成计划')
    decision=choose(config,runtime,plan['request'],mode='generate',provider=tool['id'],grants=grants)
    if decision['status']!='host_action_required':raise c.ConfigError('当前配置不再允许接收该生成任务:'+decision['status'])
    meta=inspect_image(file)
    expected={'status':'success','plan_id':plan['plan_id'],'session_id':session,'tool_id':tool['id'],'output_sha256':meta['sha256']}
    if any(receipt.get(k)!=v for k,v in expected.items()):raise c.ConfigError('回执未成功或与计划/文件不一致')
    req=plan['request']
    if not meta['has_visible_pixels']:raise c.ConfigError('生成图片全透明,缺少可见内容')
    if meta['width']<req['min_width'] or meta['height']<req['min_height']:
        raise c.ConfigError('图片尺寸小于任务要求')
    if req['transparent_background'] and not meta['has_transparency']:
        raise c.ConfigError('要求透明背景,但产物没有实际透明像素')
    meta.update(generated=True,tool_id=tool['id'],host=runtime['host'],plan_id=plan['plan_id'],
                prompt_sha256=hashlib.sha256(req['prompt'].encode()).hexdigest(),receipt_sha256=sha(receipt_path),
                provenance='agent_reported_tool_call',source_proof='回执关联已检查,非平台签名来源认证')
    return import_artifact(file,out,meta)


def import_artifact(file,out,meta):
    out=Path(out);sidecar=out.with_name(out.name+'.json')
    suffix={'.png':'png','.jpg':'jpeg','.jpeg':'jpeg','.webp':'webp'}.get(out.suffix.lower())
    if suffix!=meta['format']:raise c.ConfigError('目标扩展名须与真实图片格式一致;此入口不转码')
    if out.exists() or sidecar.exists():raise c.ConfigError('目标图片或元数据已存在,请使用新版本文件名')
    payload=Path(file).read_bytes()
    if hashlib.sha256(payload).hexdigest()!=meta['sha256']:raise c.ConfigError('验收期间图片已改变')
    meta=dict(meta,path=out.name)
    # 配对写入由协作脚本锁串行;不覆盖用户现有版本。
    with c.locked(out):
        if out.exists() or sidecar.exists():raise c.ConfigError('目标已被其他进程创建')
        no_overwrite(out,payload)
        no_overwrite(sidecar,(json.dumps(meta,ensure_ascii=False,indent=2)+'\n').encode())
    return {'status':'complete','output':str(out),'metadata':str(sidecar),'sha256':meta['sha256']}


def main():
    p=argparse.ArgumentParser(description=__doc__);sub=p.add_subparsers(dest='command',required=True)
    for name in ('image-status','image-plan','accept-image','use-image'):
        q=sub.add_parser(name);q.add_argument('--config');q.add_argument('--project')
        if name!='use-image':
            q.add_argument('--host-tools');q.add_argument('--session')
            for key in ('image-generation','cloud','paid-api','reference-upload','unknown-image-cost'):
                q.add_argument('--allow-'+key,action='store_true',help='仅表达已有的本次授权,不覆盖deny')
        if name in ('image-status','image-plan'):
            q.add_argument('--mode',choices=['off','existing','generate']);q.add_argument('--provider')
            q.add_argument('--request',required=name=='image-plan')
        if name in ('image-plan','accept-image','use-image'):q.add_argument('--out',required=True)
        if name in ('accept-image','use-image'):q.add_argument('--file',required=True)
        if name=='accept-image':q.add_argument('--plan',required=True);q.add_argument('--receipt',required=True)
    a=p.parse_args();config,_=c.effective(a.config,a.project)
    grants=[key for key in ('image_generation','cloud','paid_api','reference_upload','unknown_image_cost') if getattr(a,'allow_'+key,False)]
    runtime=runtime_tools(a.host_tools,a.session) if getattr(a,'host_tools',None) else None
    if a.command=='use-image':
        meta=inspect_image(a.file);meta.update(generated=False,mode='existing',provenance='provided_file')
        result=import_artifact(a.file,a.out,meta)
    elif a.command=='accept-image':
        if not runtime:raise c.ConfigError('需要本次最新宿主工具快照')
        result=accept(config,a.plan,a.receipt,a.file,a.out,runtime,a.session,grants)
    else:
        request=request_data(a.request) if a.request else None
        if a.command=='image-status':result=choose(config,runtime,request,a.mode,a.provider,grants)
        else:
            if not runtime:raise c.ConfigError('需要本次宿主工具快照,不能猜测工具')
            result=make_plan(config,request,runtime,a.session,a.out,a.mode,a.provider,grants)
    print(json.dumps(result,ensure_ascii=False,indent=2))
    if a.command=='image-plan' and result['status']!='host_action_required':raise SystemExit(2)


if __name__=='__main__':
    try:main()
    except (c.ConfigError,OSError,ValueError,KeyError,TypeError) as e:
        print(json.dumps({'status':'error','message':str(e) if isinstance(e,c.ConfigError) else type(e).__name__},ensure_ascii=False),file=sys.stderr)
        raise SystemExit(2)