Ukusheshisa ukuqashelwa kwebhizinisi ngokwezifiso ngokusetshenziswa kwethuluzi le-Claude ku-Amazon Bedrock

0 3 7 minutes read

Ukusheshisa ukuqashelwa kwebhizinisi ngokwezifiso ngokusetshenziswa kwethuluzi le-Claude ku-Amazon Bedrock

Amabhizinisi kuzo zonke izimboni abhekene nenselelo efanayo: indlela yokukhipha ngempumelelo ulwazi olubalulekile enanini elikhulu ledatha engahlelekile. Izindlela zendabuko zivame ukubandakanya izinqubo zokusebenzisa izinsiza kanye namamodeli angenakuguquguquka. Lokhu okuthunyelwe kwethula isisombululo esishintsha umdlalo: Ukusetshenziswa kwe-Claude Tool ku-Amazon Bedrock esebenzisa amandla amamodeli ezilimi ezinkulu (LLMs) ukuze enze ukuqashelwa okuguquguqukayo, okuguquguqukayo kwebhizinisi ngaphandle kokusetha okubanzi noma ukuqeqeshwa.

Kulokhu okuthunyelwe, sidlula:

Iyini ithuluzi le-Claude (ukushaya umsebenzi) nokuthi lisebenza kanjani
Isetshenziswa kanjani i-Claude Tool ukuze kukhishwe idatha ehlelekile kusetshenziswa imiyalo yolimi lwemvelo
Setha ipayipi elingenasiphakeli nge-Amazon Bedrock, i-AWS Lambda, ne-Amazon Simple Storage Service (S3)
Sebenzisa ukukhishwa kwebhizinisi okuguquguqukayo ezinhlotsheni zamadokhumenti ahlukahlukene
Sebenzisa isixazululo esilungele ukukhiqiza ngokulandela imikhuba ehamba phambili ye-AWS

Iyini ukusetshenziswa kwethuluzi le-Claude (ukushaya umsebenzi)?

Ukusetshenziswa kwe-Claude Tool, okwaziwa nangokuthi ukushaya komsebenzi, kuyikhono elinamandla elisivumela ukuthi sikhulise amakhono ka-Claude ngokusungula nokubiza imisebenzi yangaphandle noma amathuluzi. Lesi sici sisenza sikwazi ukuhlinzeka u-Claude ngeqoqo lamathuluzi asungulwe ngaphambili angawafinyelela futhi awasebenzise njengoba kudingeka, athuthukise ukusebenza kwawo.

Ukusebenzisa ithuluzi leClaude kusebenza kanjani ne-Amazon Bedrock

I-Amazon Bedrock iyisevisi ephethwe ngokugcwele ekhiqiza ubuhlakani bokwenziwa (AI) ehlinzeka ngezinhlobonhlobo zamamodeli esisekelo asebenza kahle kakhulu (ama-FM) avela kubaholi bezimboni abafana ne-Anthropic. I-Amazon Bedrock yenza ukusebenzisa Ithuluzi likaClaude kube lula ngokumangazayo:

Abasebenzisi bachaza isethi yamathuluzi, okuhlanganisa amagama abo, izikimu zokufaka, nezincazelo.
Kunikezwe ukwaziswa komsebenzisi okungase kudinge ukusetshenziswa kwethuluzi elilodwa noma amaningi.
U-Claude uhlola ukwaziswa futhi anqume ukuthi noma yimaphi amathuluzi angasiza ekuphenduleni umbuzo womsebenzisi noma umsebenzi.
Uma kufanele, u-Claude ukhetha ukuthi yimaphi amathuluzi azosetshenziswa nokuthi ngakuphi okokufaka.

Uhlolojikelele lwesixazululo

Kulokhu okuthunyelwe, sibonisa indlela yokukhipha izinkambu zangokwezifiso kumalayisensi okushayela kusetshenziswa i-Claude Tool ukusetshenziswa e-Amazon Bedrock. Lesi sixazululo esingenasiphakeli sicubungula amadokhumenti ngesikhathi sangempela, sikhipha ulwazi olufana namagama, izinsuku, namakheli ngaphandle kokuqeqeshwa okuyimodeli evamile.

Izakhiwo

Isixazululo sethu sokuqashelwa kwebhizinisi ngokwezifiso sisebenzisa isakhiwo esingenasiphakeli ukuze sicubungule amadokhumenti kahle futhi sikhiphe ulwazi olufanele sisebenzisa imodeli kaClaude ka-Amazon Bedrock. Le ndlela inciphisa isidingo sokuphathwa kwengqalasizinda eyinkimbinkimbi kuyilapho ihlinzeka ngamakhono okucubungula adingekayo.

Isakhiwo sesixazululo sisebenzisa izinsiza ezimbalwa ze-AWS ukuze kwakhe ipayipi elingenamthungo. Nansi indlela inqubo esebenza ngayo:

Abasebenzisi balayisha imibhalo ku-Amazon S3 ukuze icutshungulwe
Isaziso somcimbi we-S3 PUT sicupha umsebenzi we-AWS Lambda
I-Lambda icubungula idokhumenti futhi iyithumele ku-Amazon Bedrock
I-Amazon Bedrock icela i-Anthropic Claude ekukhishweni kwebhizinisi
Imiphumela ingene ku-Amazon CloudWatch ukuze igadwe

Umdwebo olandelayo ubonisa ukuthi lezi zinsizakalo zisebenza kanjani ndawonye:

Izingxenye zezakhiwo

I-Amazon S3: Igcina imibhalo yokufaka
AWS Lambda: Icupha ekulayishweni kwefayela, ithumela ukwaziswa nedatha ku-Claude, igcina imiphumela
I-Amazon Bedrock (Claude): Icubungula okokufaka nokukhipha amabhizinisi
I-Amazon CloudWatch: Ukuqapha nokungena kokusebenza kokuhamba komsebenzi

Okudingekayo

Umhlahlandlela wokuqalisa wesinyathelo ngesinyathelo:

Lo mhlahlandlela wokuqalisa ubonisa ukuthi ungakha kanjani isisombululo sokucubungula idokhumenti engenasiphakeli usebenzisa i-Amazon Bedrock kanye nezinsizakalo ezihlobene ze-AWS. Ngokulandela lezi zinyathelo, ungakha isistimu ekhipha ngokuzenzakalelayo ulwazi kumadokhumenti afana namalayisense okushayela, ukugwema ukufaka idatha mathupha kanye nokunciphisa isikhathi sokucubungula. Kungakhathaliseki ukuthi uphatha amadokhumenti ambalwa noma izinkulungwane, lesi sixazululo singakhula ngokuzenzakalelayo ukuze sihlangabezane nezidingo zakho kuyilapho sigcina ukunemba okungaguquki ekukhishweni kwedatha.

Ukusetha Indawo Yakho (imizuzu eyi-10)
1. Dala ibhakede le-S3 lomthombo wokufaka (isibonelo, okokufaka kwelayisensi yokushayela).
2. Lungiselela izindima nezimvume ze-IAM:

{
     "Version": "2012-10-17",
     "Statement": [
       {
         "Effect": "Allow",
         "Action": "bedrock:InvokeModel",
         "Resource": "arn:aws:bedrock:*::foundation-model/*", "arn:aws:bedrock:*:111122223333:inference-profile/*”
       },
       {
         "Effect": "Allow",
         "Action": "s3:GetObject",
         "Resource": "arn:aws:s3:::amzn-s3-demo-bucket/*"
       }
     ]
   }

Ukudala umsebenzi we-Lambda (imizuzu engama-30)

Lo msebenzi we-Lambda ucushwa ngokuzenzakalelayo lapho isithombe esisha silayishwa ebhakedeni lakho le-S3. Ifunda isithombe, isifake ngekhodi ku-base64, bese isithumela ku-Claude 4.5 Sonnet nge-Amazon Bedrock isebenzisa i-API yokusebenzisa Ithuluzi.Umsebenzi uchaza ithuluzi elilodwa elibizwa ngokuthi. khipha_izinkambu_zelayisensi ngezinhloso zokubonisa. Kodwa-ke, ungachaza amagama amathuluzi nezikimu ngokusekelwe esimweni sakho sokusebenzisa — isibonelo, ukukhipha idatha yekhadi lomshwalense, amabheji omazisi, noma amafomu ebhizinisi. U-Claude ukhetha ngokuguqukayo ukuthi alishayele yini ithuluzi lakho ngokususelwe ekuhlobaneni okusheshayo nesakhiwo sokufakwayo.

Sisebenzisa “tool_choice”: “okuzenzakalelayo” ukuvumela u-Claude anqume ukuthi uzowucela nini umsebenzi. Ezimweni zokusetshenziswa kokukhiqiza, ungase ufune ukufaka ikhodi eqinile “tool_choice”: {“type”: “ithuluzi”, “name”: “your_tool_name” } ngokuziphatha okunqunyiwe.

Iya ku-AWS Lambda console
- Khetha Dala umsebenzi.
- Khetha Umbhali kusukela ekuqaleni.
- Setha isikhathi sokusebenza sibe I-Python 3.12.
- Khetha Dala Umsebenzi.
Lungiselela i-Lambda Timeout
- Ekucushweni kwakho komsebenzi we-Lambda, chofoza Ukucushwa Okujwayelekile ithebhu.
- Ngaphansi Ukucushwa Okujwayelekilechofoza Hlela
- Ngoba Isikhathi siphelilekhuphuka usuke kumasekhondi angu-3 azenzakalelayo uye okungenani amasekhondi angu-30. Sincoma ukuthi uyisethe kumaminithi angu-1-2 ukuze uthole izithombe ezinkulu.
- Khetha Londoloza.
  
  Qaphela:
  Lokhu kulungiswa kubalulekile ngoba ukucubungula izithombe ngo-Claude kungase kuthathe isikhathi eside kunesikhathi esimisiwe se-Lambda, ikakhulukazi ezithombeni ezinokulungiswa okuphezulu noma lapho kucutshungulwa izinkambu eziningi. Gada isikhathi sokwenza komsebenzi wakho kokuthi CloudWatch Logs ukuze ulungise lesi silungiselelo esimweni sakho esithile sokusebenzisa.

Namathisela le khodi ku- lambda_function.py ifayela lekhodi:

import boto3, json
import base64

def lambda_handler(event, context):
    bedrock = boto3.client("bedrock-runtime")
    s3 = boto3.client("s3")
    
    bucket = event["Records"][0]["s3"]["bucket"]["name"]
    key = event["Records"][0]["s3"]["object"]["key"]
    file = s3.get_object(Bucket=bucket, Key=key)
    
    # Convert image to base64
    image_data = file["Body"].read()
    base64_image = base64.b64encode(image_data).decode('utf-8')
    
    # Define tool schema
    tools = [{
        "name": "extract_license_fields",
        "input_schema": {
            "type": "object",
            "properties": {
                "first_name": { "type": "string" },
                "last_name": { "type": "string" },
                "issue_date": { "type": "string" },
                "license_number": { "type": "string" },
                "address": {
                    "type": "object",
                    "properties": {
                        "street": { "type": "string" },
                        "city": { "type": "string" },
                        "state": { "type": "string" },
                        "zip": { "type": "string" }
                    }
                }
            },
            "required": ["first_name", "last_name", "issue_date", "license_number", "address"]
        }
    }]
    
    payload = {
        "anthropic_version": "bedrock-2023-05-31",
        "max_tokens": 2048,
        "messages": [{
            "role": "user",
            "content": [
                {
                    "type": "image",
                    "source": {
                        "type": "base64",
                        "media_type": "image/jpeg",
                        "data": base64_image
                    }
                },
                {
                    "type": "text",
                    "text": "Extract the driver's license fields from this image."
                }
            ]
        }],
        "tools": tools
    }
    
    try:
        response = bedrock.invoke_model(
            modelId="global.anthropic.claude-sonnet-4-5-20250929-v1:0",
            body=json.dumps(payload)
        )
        
        result = json.loads(response["body"].read())
        
        # Print every step for debugging
        print("1. Raw Response:", json.dumps(result, indent=2))
        
        if "content" in result:
            print("2. Content found in response")
            for content in result["content"]:
                print("3. Content item:", json.dumps(content, indent=2))
                
                if isinstance(content, dict):
                    print("4. Content type:", content.get("type"))
                    
                    if content.get("type") == "text":
                        print("5. Text content:", content.get("text"))
                    
                    if content.get("type") == "tool_calls":
                        print("6. Tool calls found")
                        extracted = json.loads(content["tool_calls"][0]["function"]["arguments"])
                        print("7. Extracted data:", json.dumps(extracted, indent=2))
        
        return {
            "statusCode": 200,
            "body": json.dumps({
                "message": "Process completed",
                "raw_response": result
            }, indent=2)
        }
        
    except Exception as e:
        print(f"Error occurred: {str(e)}")
        return {
            "statusCode": 500,
            "body": json.dumps({
                "error": str(e),
                "type": str(type(e))
            })
        }

Hambisa Umsebenzi we-Lambda: Ngemva kokunamathisela ikhodi, khetha okuthi Sebenzisa inkinobho ohlangothini lwesobunxele lomhleli wekhodi bese ulinda umlayezo wokuqinisekisa ukuthunyelwa.
Okubalulekile: Khumbula njalo ukusebenzisa ikhodi yakho ngemva kokwenza izinguquko. Lokhu kuqinisekisa ukuthi ikhodi yakho yakamuva igciniwe futhi izosetshenziswa uma umsebenzi we-Lambda uqaliswa.

Ukusebenza ne-Claude Tool sebenzisa izikimu
1. I-Amazon Bedrock ene-Claude 4.5 Sonnet isekela umsebenzi wokushaya kusetshenziswa ukusetshenziswa kweThuluzi – lapho uchaza khona amathuluzi ashayekayo ngezikimu ezicacile ze-JSON. Ithuluzi elisebenzayo elifakiwe kufanele lifake:
  - igama: Inkomba yethuluzi lakho (isb extract_license_fields)
  - input_schema: I-schema se-JSON esichaza izinkambu ezidingekayo, izinhlobo, nesakhiwo
2. Incazelo yokusebenzisa Ithuluzi eliyisibonelo:
```
[{
  "name": "extract_license_fields",
  "input_schema": {
    "type": "object",
    "properties": {
      "first_name": { "type": "string" },
      "last_name": { "type": "string" },
      "issue_date": { "type": "string" },
      "license_number": { "type": "string" },
      "address": {
        "type": "object",
        "properties": {
          "street": { "type": "string" },
          "city": { "type": "string" },
          "state": { "type": "string" },
          "zip": { "type": "string" }
        }
      }
    },
    "required": ["first_name", "last_name", "issue_date", "license_number", "address"]
  }
}]
```
3. Ungachaza amathuluzi amaningi ku- uhlu lwamathuluzi. UClaude ukhetha eyodwa (noma akekho) kuye ngokuthi ithuluzi_lokukhetha inani nokuthi ukwaziswa kufana kanjani kahle ne-schema esinikeziwe.
Lungiselela Isaziso Somcimbi we-S3 (imizuzu emi-5)
1. Vula ikhonsoli ye-Amazon S3.
  - Khetha ibhakede lakho le-S3.
  - Chofoza u- Izakhiwo ithebhu.
  - Skrolela phansi ukuze Izaziso zomcimbi.
  - Chofoza Dala isaziso somcimbi.
  - Faka igama lesaziso (isb, “LambdaTrigger”).
  - Ngaphansi Izinhlobo zomcimbikhetha PUT.
  - Ngaphansi Indawo oya kuyokhetha Umsebenzi we-Lambda.
  - Khetha umsebenzi wakho we-Lambda kusukela kokwehlayo.
  - Chofoza Londoloza izinguquko.
Ukuhlola Nokuqinisekisa (imizuzu eyi-15)
1. Amafomethi Asekelwe: I-Claude 4.5 isekela okokufaka kwesithombe ku-JPEG, PNG, WebP, namafomethi we-GIF yohlaka olulodwa. Qaphela: Nakuba lokhu kuqaliswa okwamanje kusekela kuphela .jpeg izithombe, ungakwazi ukunweba ukwesekwa kwamanye amafomethi ngokulungisa ifayela le- uhlobo_lwemidiya inkambu kumsebenzi we-Lambda ukuze ifane nohlobo lwefayela elilayishiwe lwe-MIME.
2. Usayizi kanye nemikhawulo yokulungiswa:
  - Usayizi wesithombe esiphezulu: 20 MB
  - Ukulungiswa okunconyiwe: 300 DPI noma ngaphezulu
  - Ubukhulu bobukhulu: 4096 x 4096 pixels
  - Izithombe ezinkulu kunalezi zingase zehluleke ukucubungula noma ukukhiqiza imiphumela engalungile.
Amathiphu Okucubungula Ngaphambili Okunemba Okungcono:
1. Sika isithombe ngokuqinile ukuze ususe umsindo nezigaba ezingabalulekile.
2. Lungisa ukugqama nokukhanya ukuze uqinisekise ukuthi umbhalo ufundeka ngokucacile.
3. Susa ukuskena futhi uqinisekise ukuthi umbhalo uqondaniswe ngokuvundlile.
4. Gwema izithombe-skrini ezinokulungiswa okuphansi noma izithombe ezinama-artifact okucindezela kanzima.
5. Uncamela ingemuva elimhlophe nombhalo omnyama ukuze uthole ukucaca okukhulu kwe-OCR.
Layisha Isithombe Sokuhlola:
1. Vula ibhakede lakho le-S3
2. Layisha isithombe selayisensi yokushayela (amafomethi asekelwayo: .jpeg, .jpg).
3. Qaphela: Qinisekisa ukuthi isithombe sicacile futhi siyafundeka ukuze uthole imiphumela ehamba phambili.
Gada Izingodo ze-CloudWatch
1. Iya kukhonsoli ye-Amazon CloudWatch.
2. Chofoza ku- Amaqembu okungena ekuzulazuleni kwesokunxele.
3. Sesha igama lomsebenzi wakho we-Lambda invoke_drivers_license.
4. Chofoza ukusakazwa kwelogi kwakamuva (kuhlungwe ngesitembu sesikhathi).
5. Buka imiphumela yokusebenzisa, ebonisa lesi sampula esiphumayo:

{ 
  "type": "tool_use",
  "id": "toolu_bdrk_01Ar6UG7BcARjqAKsiSPyNdf",
  "name": "extract_license_fields", 
  "input": { 
        "first_name": "JANE",
        "last_name": "DOE", 
        "issue_date": "05/05/2025", 
        "license_number": "111222333", 
        "address": { 
            "street": "123 ANYWHERE STREET", 
            "city": "EXAMPLE CITY", 
            "state": "VA", 
            "zip": "00000"
               } 
           }
 }

Ukuthuthukisa ukusebenza

Lungiselela inkumbulo ye-Lambda nezilungiselelo zokuvala isikhathi
Sebenzisa ukucubungula iqoqo kumadokhumenti amaningi
Sebenzisa izaziso zomcimbi we-S3 ngokucubungula okuzenzakalelayo
Engeza amamethrikhi e-CloudWatch ukuze uqashwe

Izindlela ezihamba phambili zokuphepha

Sebenzisa ukubethela lapho uphumule amabhakede e-S3
Sebenzisa okhiye be-AWS Key Management Service (KMS) ukuze uthole idatha ebucayi
Sebenzisa izinqubomgomo ze-IAM ezinelungelo elincane
Nika amandla iziphetho zefu eliyimfihlo (i-VPC) ukuze uthole ukufinyelela kwenethiwekhi yangasese

Ukuphatha iphutha nokuqapha

Okuphumayo kukaClaude kuhlelwe njengohlu lwamabhulokhi wokuqukethwe, okungase kuhlanganise izimpendulo zombhalo, ithuluzi_izingcingonoma ezinye izinhlobo zedatha. Ukulungisa iphutha:
1. Ngaso sonke isikhathi bhala impendulo eluhlaza evela ku-Claude.
2. Hlola uma tool_calls ikhona empendulweni.
3. Sebenzisa i-try-ngaphandle block eduze kwekholi yomsebenzi ukuze ubambe amaphutha afana nokulayisha okungalungile noma ukuphela kwesikhathi kwemodeli.
Nali iphethini yokuphatha iphutha elincane:

try:
    result = json.loads(response["body"].read())
    if "tool_calls" in result.get("content", [{}])[0]:
        args = result["content"][0]["tool_calls"][0]["function"]["arguments"]
        print("Extracted Fields:", json.dumps(json.loads(args), indent=2))
except Exception as e:
    print("Error occurred:", str(e))

Hlanza

Susa ibhakede le-S3 nokuqukethwe.
Susa imisebenzi ye-Lambda.
Susa izindima nezinqubomgomo ze-IAM.
Khubaza ukufinyelela kwe-Bedrock uma kungasadingeki.

Isiphetho

Ukusetshenziswa kwe-Claude Tool ku-Amazon Bedrock kunikeza isisombululo esinamandla sokukhipha inkambiso yangokwezifiso, ukunciphisa isidingo samamodeli ayinkimbinkimbi okufunda ngomshini (ML). Lesi sakhiwo esingenasiphakeli sinika amandla ukucutshungulwa kwamadokhumenti okulinganiswayo, okungabizi kakhulu ngokusethwa nokunakekelwa okuncane. Ngokusebenzisa amandla amamodeli ezilimi ezinkulu nge-Amazon Bedrock, izinhlangano zingavula amazinga amasha okusebenza kahle, ukuqonda, kanye nokusungula izinto ezintsha ekuphatheni idatha engahlelekile.

Izinyathelo ezilandelayo

Sikukhuthaza ukuthi uhlole lesi sixazululo ngokuqhubekayo ngokusebenzisa ikhodi yesampula endaweni yakho futhi uyenze ngendlela oyifisayo izimo zakho zokusebenzisa ezithile. Joyina ingxoxo mayelana nezixazululo zokukhipha ibhizinisi kumphakathi we-AWS re:Post, lapho ungabelana ngomuzwa wakho futhi ufunde kwabanye onjiniyela.

Ukuze uthole imininingwane ejulile yobuchwepheshe, hlola imibhalo yethu ebanzi ku-Amazon Bedrock, AWS Lambda, kanye ne-Amazon S3. Cabangela ukuthuthukisa ukusebenzisa kwakho ngokuhlanganisa ne-Amazon Textract ukuze uthole izici ezengeziwe zokucubungula amadokhumenti noma i-Amazon Comprehend ukuze uthole ukuhlaziywa kombhalo okuthuthukile. Ukuze uhlale unolwazi ngezixazululo ezifanayo, bhalisa ku-AWS Machine Learning Blog yethu futhi uhlole izibonelo ezengeziwe endaweni ye-AWS Samples GitHub. Uma umusha ezinsizeni zokufunda zomshini we-AWS, hlola i-AWS Machine Learning University yethu noma uhlole Ilabhulali yethu Yezisombululo ze-AWS. Ukuze uthole izixazululo nosekelo lwebhizinisi, finyelela ngethimba lakho le-akhawunti ye-AWS.