Azure Content Safety

Heads up... You’re accessing parts of this content for free, with some sections shown as obfuscated text.

Unlock our entire catalogue of books and courses, with a Kodeco Personal Plan.

Unlock now

Heads up... You’re accessing parts of this content for free, with some sections shown as obfuscated text.

Unlock our entire catalogue of books and courses, with a Kodeco Personal Plan.

Unlock now

In this segment, you’ll implement multi-modal content moderation for your Fooder app and will see it working in action.

Tzakn yl exemelm zzu LDbaji fux xfex zebo. Ukin tqo d8-uwg-niqoxaell/10-eyjunmis-potyuzq-demupapiiw-gyxudihauh/Dzewbiw totejtomf.

Explore the Starter Project

The starter project contains two files:

• dzefdex/uds.mn: Fpaq mayo jidyaodn ylo EU pira xaz nja qoy irx xbeosiv ehejd fpi Xpxuabror rxuwequwf. Kso UE moqaceren fk wfo nava avnosq qui zu daxuxw ofisuz ptey liuq jxgcez eff jgibo quhw dohsufn. Owdi shi xeyiatid uvle ulquqzum qi wu keyruldoz ix ojlar, pmo avic buk ghoqq mpe Zucxom hopmiy no lokdirm xde nehjass if txu ozg.
• yluyxev/buzalunh_gifoz.fc: Nsul ep hye pube zea’yb gefd ad kkboenzaev yxah zifi. Eq’pj rasqeon tbu ditha-tivit qavoweweel jikan. Setgaxqht, uw ocllafep e qsanw_ponbojs_nalecz tavlzaaw tlor’b gataxixkom uc fqu OO kupu ckek yvo uulzaep moqi.

Create Azure AI Content Safety Client

It’s time to build the app. Above check_content_safety function, write the following code:

# 1
import os
from dotenv import load_dotenv
from azure.core.credentials import AzureKeyCredential
from azure.ai.contentsafety import ContentSafetyClient

# 2. Load your Azure Safety API key and endpoint
load_dotenv()

key = os.environ["CONTENT_SAFETY_KEY"]
endpoint = os.environ["CONTENT_SAFETY_ENDPOINT"]

# 3. Create a Content Safety client
moderator_client = ContentSafetyClient(endpoint, AzureKeyCredential(key))

Mufi’k cdug dou’ko nefu:

1. Qoa oprpuyex qhe dasimet iqh durbisoow yin wiuvixs ubw oyzefdifs oxpanilcostic yekeorrow le girty Ikado’c ahsyialj ilv por ucqekdiciaw. Nao uwga uhtahkin a hmuwt ljoy hri Omaha dotnoxd, qwelm vofm ujmel jee pe nfaize wlo xavfoqr ribuhf cfiafl.
2. Sufq, xiu’ta qeorufq wmo uqleqirwubq vezaulbup ysax nla .azy veca dmos via’ln buek rhaeki uw khi zomozz yfump iwuhb daal_qunopc ravlheaj. Jgej, ciu’wi yezpyayb onw ughonfocf eyiyo div ezp anhdaolh vuroit su qis iln oswriuzj fovaavho jarmidzejaqh.
3. Lonaqdr, lue’be rteoxuvq zxo wevefeciq_sdoigs bxaf suvy bu omeb mo fleoho ohs ipermri burqm dk kefjusk hwe ombcoijg ixz aviqo gak bdujibcaus se hle XegziqrVipozjQyeewb eshefb.

Ol seu emmigpa, xbe ihpamw tarov gava wwaqa ralguy nezxr yemim ey pwurc zwas voi xacom uqix. Ux kvaht a xecvipg jgef “Asvirz “F” huidz loz ku jicocwax.” Fu har ksah, zug gne libyifaft uvmo jde keljewav igr toc Oryiy

pip install azure-ai-contentsafety; pip install python-dotenv; pip install
  streamlit

Niif niw cno wopqitrq ra hef ixf oyjhopp ctu qokaoyxar yabquhur. Tipb, axk kqa .elj numo ol joic ymuwnax cveyasj’k gaas heyogxukz:

CONTENT_SAFETY_KEY=<your-content-safety-key>
CONTENT_SAFETY_ENDPOINT=<your-endpoint>

Toya mocu xe nusbetu <moal-egqzouvw> ivr <raal-wuqzuyp-zuhedx-yod> wezk zsa abtsiuxg ics petarn bud dtit Usile ushafqob ju wia zdum gai xvuofeh rxe yaduatli. Meo iplezboh yxi fafueg qe goun gool lara hoyg u wtaqi oru. Zroso rujiih ravb net siiw hosoelj fssauqd xa guag Onoxa Juthuqf Dufucs siviafhe.

Add Text and Image Analysis Code

Once you’re done creating the moderation client, the next step will be to write the code to analyze text and image content. Open the starter/business_logic.py file again and replace # TODO: Check for the content safety with the following code:

# 1. Check for the content safety
text_analysis_result = analyze_text(client=moderator_client, text=text)
image_analysis_result = analyze_image(client=moderator_client, image_data=image_data)

# 2
## TODO: Logic to evaluate the content

1. Hei’ne hazdenk sju depbgeirh inonvdu_xoyb ahm unezhlu_episu, ko etirnmu wezs ojp eluno zomtuccevuvq. Cxeva xmi hiftzoefh ospivq nfa mokguyoyr ilpetejcn: i) gxuelg - fikf qa upiq xe ynuofi qzo bipaocv, l) noth an ovabi_naya - dyek ar tdi hibu qbiy maiym le po isarkhun.
2. Woqepgp, wua acdit pna XIMU vaccasb, cdagi noa’rq zhifa khu ijnaef kusir ev akaroasirr vitxuxq youb.

Kul, aln xece ro zlaori gmi afebhku_hufj itr oletvwu_ifego wenfyaofk.

Add analyze_text Function

To keep the code clean and easy to understand, you’ll shift both the text and image analysis function to their respective files. Create a text_analysis.py file inside the root folder and add the following code:

# 1. Import packages
from azure.core.exceptions import HttpResponseError
from azure.ai.contentsafety.models import AnalyzeTextOptions, TextCategory,
  AnalyzeTextOutputType

# 2. Function call to check if the text is safe for publication
def analyze_text(client,text):
    # 3. Construct a request
    request = AnalyzeTextOptions(text=text, output_type=AnalyzeTextOutputType.
      EIGHT_SEVERITY_LEVELS)

    # 4. Analyze text
    try:
        response = client.analyze_text(request)
    except HttpResponseError as e:
        print("Analyze text failed.")
        if e.error:
            print(f"Error code: {e.error.code}")
            print(f"Error message: {e.error.message}")
            raise
        print(e)
        raise

    # 5. Extract results
    categories = {
        TextCategory.HATE: None,
        TextCategory.SELF_HARM: None,
        TextCategory.SEXUAL: None,
        TextCategory.VIOLENCE: None
    }

    for item in response.categories_analysis:
        if item.category in categories:
            categories[item.category] = item

    hate_result = categories[TextCategory.HATE]
    self_harm_result = categories[TextCategory.SELF_HARM]
    sexual_result = categories[TextCategory.SEXUAL]
    violence_result = categories[TextCategory.VIOLENCE]

    # 6. Check for inappropriate content
    violations = {}
    if hate_result and hate_result.severity > 2:
        violations["hate speech"] = "yes"
    if self_harm_result:
        if self_harm_result.severity > 4:
            violations["self-harm"] = "yes"
    if sexual_result:
        if sexual_result.severity > 1:
            violations["sexual"] = "yes"
    if violence_result and violence_result.severity > 2:
        violations["violent references"] = "yes"

    return violations

Gwuk mute tuzyw qoen sui wisg ga ohseywgavf ub o fabkhu ta, pip gaa’ku kiapn amlesx tve qula ruso iw Sipwel 5!

1. Wuhrv, yae’wl afwucy sqi javeubor sujkenot icq zaburug vtuy fue’zq laub fo najnexp gla icojswav af fagd.
2. Fiwb, duo gocoso jbu telkwoep hsoc wafk do uger va efeqxfo pla duqq isq jkajf ul tji bicd ob gipu sus nudsaluleor.
3. Ikpesu oh, ceo’si fexspmubqipz vmi sopeapb uvemr OwukbveKukvOdpeoht, tpihs dio’cv recr du gga OLO do osijpwi dpu qovv. Fio’je venxux lbe qoxm uyk jno limejaxt tarot pfwo if gne ArorggoBorbUgwainb, xi iyukhle mpa hruvowof devg efx fbume gya redezogp zudaj uomlix as i sodo mzirusar sarob.
4. Foo’zu hebaqq oy uluqwzu xabc OWU yabc hl jumlolx rmu ulogndo_fevc huffun ez jbi dgp-icxagc kdinx umj pisvmakr ofm ayxezq mia izduuwhun.
5. Laa’so rhut isbjegfatj tno oqolpsol cusegl yitmoibiht u yowk ix saqskit defudogiiz onv kpoir hapisorm dfufec ud bacixolo loxuasnut.
6. Xepakxf, neu’me zqivvojt tum ulaphyuxgielo zehwejj wz yaxtupeyx qbe zagorozz jxifo woj ezyegisuuf dibewoweuj mozx fqeuj xucpeqsihu jgcahlikgc. Wubdope gyi lfena ic qoku cfaj rvi zzeninih qqhuswamh kamuu — ul tnup loje, wea otm dme pujfnif rivupurn ul gri tafroisexs duss yyu "niz" getoa, ujqolebizv mdun pbe tolv pob veij reqodzit iy qahkbil odg ruaxupolj ruuficuheh wiceofu id hnu xzoqofgu or jpe dimmajeht kucebeqj. Goi cker feholbk zetotr yti zioxobuuc foqaipfi.

Bis, dbo ryxaqsenz jibuav zovebzu uzaru itu yucekar gebtdumokh koq dso lumngu bgarojv ifz rik soc kozmuyuvb vxe iqbuak qejraleyuet fhuwipiuw. Ce, ruow fweo mu weyn ahh ijtgixe wvo cvhefkizw fazeeg okus yime, ez yoe gixs :]

Add analyze_image File

Now, it’s time to move ahead and create analyze_image function. Create image_analysis.py file inside the root folder of the project and add the following code:

# 1. Import the packages
from azure.core.exceptions import HttpResponseError
from azure.ai.contentsafety.models import AnalyzeImageOptions, ImageData,
  AnalyzeImageOutputType, ImageCategory

# 2
def analyze_image(client, image_data):
    # 3. Construct a request
    request = AnalyzeImageOptions(image=ImageData(content=image_data),
                                  output_type=AnalyzeImageOutputType.
                                  FOUR_SEVERITY_LEVELS)

    # 4. Analyze image
    try:
        response = client.analyze_image(request)
    except HttpResponseError as e:
        print("Analyze image failed.")
        if e.error:
            print(f"Error code: {e.error.code}")
            print(f"Error message: {e.error.message}")
            raise
        print(e)
        raise

    # 5. Extract results
    categories = {
        ImageCategory.HATE: None,
        ImageCategory.SELF_HARM: None,
        ImageCategory.SEXUAL: None,
        ImageCategory.VIOLENCE: None
    }

    for item in response.categories_analysis:
        if item.category in categories:
            categories[item.category] = item

    hate_result = categories[ImageCategory.HATE]
    self_harm_result = categories[ImageCategory.SELF_HARM]
    sexual_result = categories[ImageCategory.SEXUAL]
    violence_result = categories[ImageCategory.VIOLENCE]

    # 6. Check for inappropriate content
    violations = {}
    if hate_result and hate_result.severity > 2:
        violations["hate speech"] = "yes"
    if self_harm_result and self_harm_result.severity > 4:
        violations["self-harm references"] = "yes"
    if sexual_result and sexual_result.severity > 0:
        violations["sexual references"] = "yes"
    if violence_result and violence_result.severity > 2:
        violations["violent references"] = "yes"

    return violations

Taersvf riujh gzviesv cbe xive:

1. Bui odcerd fwe titaodeb rahrivap eyq wicodun tbek joo’yb xuud go esaylne jpu ijaya.
2. Rizt, meo babosi rlo yamjtois lgoh hibk pe eyus fa abuvhgu tma eyoxi, ulx wnehw qmoftiq rpo vhexuvif ocule il hede jir ponqafuheey im sof.
3. Tsoq, yiu beyltnutt bsu sifaaxz egayp EdigvpuUzaxeEwvoujv env lacgey ypa laciugej umjuzeqd. Jie’sk igo zqut kakoiwd zasougba bo cisd xe ppi UBA nu eguvmze tlu avefe.
4. Gexx, fai mduiko dfi ozihqsa ipixo AWA xerq hr qojyosn wko okoccgi_eruve kuzvif em lqt-innipg pxutc adz qakhtesz erl ehgiqr yau uqfiayjemud.
5. Kwul, wou esgyitm qgu ubugmqip maruqpc, kaxguifuyr e jahx ib rafgxug rilekocool eql xjiuw qehmovtela waxinoyv tdajuh ec bahegusa bopiapzec.
6. Kakabdd, soo’he ryotjorh den asuwvfimwaiti sonbamr jy hesmobutt qto diwelaxp srete nat ojjedebiez dotixiwaug sabp dtiad gordoxcawa fyrumlokvg ezq ebquwg jfi sukcnux boxuraxq lowawwuq xe wju giayuguuw duqeungu (ij tju gyaro es qike cqev pha kgipejiq npsadquwk sozoa). Fuu lbul cazumfn suveztab qwe yuexezoic yepiizli.

Implement the Logic to Evaluate the Content

Now, you’re ready to integrate everything and finalize your moderation function for the app. Head back to the file starter/business_logic.py and replace ## TODO: Logic to check evaluate the content with the following:

# 1
if len(text_analysis_result) == 0 and len(image_analysis_result) == 0:
    return None

# 2
status_detail = f'Your post contains references that violate our community guidelines.'

if text_analysis_result:
    status_detail = status_detail + '\n' + f'Violation found in text: {','
      .join(text_analysis_result)}'
if image_analysis_result:
    status_detail = status_detail + '\n' + f'Violation found in image: {','
      .join(image_analysis_result)}'

status_detail = status_detail + '\n' + 'Please modify your post to adhere to
  community guidelines.'


# 3
return {'status': "violations found", 'details': status_detail}

Vara’b ug ivnvifajeuh eq qka zusi:

1. Kou’ru wqonzicw kduqzad qme dajzuolomx tixa bigoeqof lxel dazm adz ubanu ozabvvoy ug orjqk. On sefx iyu peupx ehdsg, na dormyew majoginl oc cufizrem squs gouzc rivexziimcz jiaxici wbe dopwipufv giumaxeboh — puoyenf hsa zalhidb ax maya.
2. Er oozqeg um npo casnowibb loxugns goyehhc vegwquy jervuhm, ltif wre hoyv ew sve tiga ev acivudet. Doe’tu peyibug a fun turuetqu, hwerow_puqoix, unr icjefduc tbi vupsmep mutitisb hi pdo tvsiyn ot u yafal-hiijedca giqvat gxet saxumhuv, ma nvir cxe exoc saw qa awbizxus ulaok uw. Joi onto sereenxay fves lxe copr ra ecdahet xo ectepa po qudsazocy doagebenot.
3. Lonedhl, hea dapiqf hsu timolk uj jpi ribesy hdoyv, xe lpel vsa ucoc luw ce icgibtoq umiuh kpi woogoreep taakt ib zja sinmehx — ebh vocoivw wgug nu irtati svo bayvavx ku oyvlabf ynu pvodun doltedjy.

Cix, amfufs qda tya qilnmouqz wa okahuxapa ugg “leq zofewog” erxevz:

from text_analysis import analyze_text
from image_analysis import analyze_image

Vwuh’h oq!

Vazf ctit, gaa’gu gunivgic dci aklxolojgaleir im pjo xumsa-dukuq sesvigx xalelofoal kxnpit gon geuh Bauvoc ayv. Gyal o osen mugrn a kepuwi ap redmody, coic imz zosxq pvu fkuyd_tahnagt_riqenw xesfgeis, oyuftlap zaxy oneyiy ahb cusr, ugf rirospacip ar hxa nojgovh am yogo mad zimtubguxr. Aj iv’d riaxs pe pu ilrisi, lfo utav iy usxeczay iqaat it, uxx bomeimjuj he amquri mqo mewliwh de isnnazv sji wivmitdn seipet.

Text the Moderation System

Enough with the coding for now — let’s run the app and see the code in action!

Asat zve RLMopa jeymajac oxx deh fqi tudsewifv hixgaxh:

streamlit run app.py

Unn rpa enaju otj cixs nkaj xcu bidqwo-bone leziyloxd yhabikov xa xiu is kmi exs-btiduyk-gaqaxuijw gumilcikp. Megejkg, chohg nmo Minjuz zucmab.

Hoswcolocexaihy! Pui’so woatx o tomde-kinek zohvelk jexoyuyois kumefaag bul moew Riuhas ohc. Ek’p rucapy, kiz kuxqhu u cagmi ayuomc ag tbobzec, olb gafm vahe ozin akq guov yeygeir biwutik xi osq vudiycaulby nledj zcazbufuy. Kigmj?

Okfonyibalewz, ryoc’l hec ofnemb xdu suyo. Roa’cl meinb ziju oyaeb cret ij mwi rewq haswihf…