Slides: Building Conversational Agents — Thor Schaeff and Philipp Schmid, Google DeepMind
Source Video
Building Conversational Agents — Thor Schaeff and Philipp Schmid, Google DeepMind
Relationship To World's Fair 2026
These slides are extracted from a public AI Engineer YouTube video connected to World's Fair 2026. Speaker-matched clips are supporting context unless later confirmed as exact session recordings; official livestream recordings are day-level/event-level source material.
Related Scheduled Sessions
- No individual scheduled session mapping has been assigned yet; treat this as an event livestream deck.
Extracted Slides
OCR text:
PLATINUM SPONSORS
Braintrust WorkOS OpenAI
OCR text:
One API to
Rule Them All
OCR text:
One API to
Rule Them All
OCR text:
Chro
WedAor813:08
Ose APi lo Rule TheA:B
taistudio.google.compromptshew_chat
GoogleAI Studio
Playground
（>Oetcode
GX
Playground
Gemma431BIT
AME2026Score
pmg-4-31b-5
gleDeepMind'sTlagshpopen-weighe
ema2310 TMo
ExploreGooglemodels
eodepba
Hteuo
256Kcontextwindowandsdv
ma
Secure Cioud Run Service Acc
Gcloud Login Wh Base64 JSON
System instructions
Featured
Code and Chat
View ali history
Image Generation
2Buld
wthGemi3
No API Key
Dashboard
Swch toa pad APikey to unlock higer
qucta aod more feetures
Documentation
Video Generation
Speech and Music
Real-time
Temperature
Explore our text to speech and music
Real-time voice and ic
Gemini Lve.
eowith
ceneration models
mode's
Media resolution
Default
-bupnqes
Thinking level
ubject to the Gemma Terms o
Use
High
QSearch
QWhat'snew
GetAPIkey
Structured outputs
③Setings
B8Tools
GGrounding withGoogle Search
ognoaxapo
schmidphllpp1995@
OCR text:
Ce i oe nee eS © 6a @ FF w@ . EF weet
° ; a / @ °
A :
Google Al Studio \
API-Schliissel 3D APL Kur zantedung fe ADL-Schtussel erstesen
Pasnboara
APESchtussel ‘
Cucpee cor © API-Schtussel Proyeict oa 10 Ale Prowate »
Projekte
Mass Scnhissel Preyeht Eestett Aten maungestute
Rotvabeqrenzur|
Auagaben
AD ee haung
Protoagiie und Datasets
Arderungupotoket *
Gre konnen Inte AM -Schiussel her nicht finden?
Proyckte importioren
+ Search ,
1. Yenat snee
+ Get APL key
Ih Setings
@ vc rerdpr ppl gonce
OCR text:
, ; OS ae
Race i . Pooh!
aos ; : J i
ahs an
OCR text:
PR eC cS cet oi oT od Gor G6 & & CB SF mw © sects
bs ‘ ad °o
‘
J
J
Al ENGINEER EUROPE - WORKSHOP
ae
Building Agents with the Gemini Interactions API, from zero toa 5 os a
working agentic application in 60 minutes.
Pk git ee aR ol ee ona
© 60 min: ive Coding =A Philipp Schmid
OCR text:
Oe Cn ee OR ee CO a . i ne ce err ers
CJ a a ry
Part 1; The Interactions APi
Core Primitives
.
a OO ———————————————————————————————————————— OR _—_—eeeeeeeeoeeee
C
, © Server-Side State * Background Execution
Be Ce eh Ma aes Ge OST a Le OPM MER Cle as ste aed
Pana)
Chain turns with peectous ieteractice <3, Server owns conversation Fire-and-forget for long tasks Poll with .-re-o:tors getiagi.
history.
Ov
(3 Typed Output Blocks © Streaming via SSE
€ irteraition cutouts raat client isterectiors
Pera Ca
oe mas ae
a i. f
ai
Flat areay: vert, totic all amage, teat?
RICN EVENS: coment aaling ierer action rome tere.
OCR text:
Help 口 Aor813:19
Code Comparison Part 1:The Interactions API
GENERATECONTENT INTERACTIONSAPI
r1·client.nodels.generate_content( Stateless:re-send history every turn contents-[ ("inline_dato”:(mime_type":imoge/jpeg” ("text":Describe this inoge") "doto：...")) i1-client.interactions.createc model-'gemini-3-flosh-preview input-[ 1x01.0d1.） fou.:od1.) "uri":"https://... 6od/ou.:odou
i2-client.interactions.creote( Request 2:Just-send new Input
r2client.nodels.generate_content( model-'genini-3-flosh-preview" contents-[ previous_interoction_id-i1.id input=*Is there any text?
("inline_doto”:("mime_type”:“imogo/jpeg” ("text”:“Describe this inoge) "doto:"..."))
("role":"model" osn,oo.) （[（oo1odo.Ix）]1d （[（22x）u0o4）sI.x.）]:s1od
Access:response.condidates[e] content.ports[e].function_coll
8/17
OCR text:
Pe en Ca eT ©: GO ek OF @ LE meters
e . a 5 °
Part 2: Building Agents
Co
. © HOW IT WORKS
: @ Send request: prompt - tool definitions erence
e@ Check outputs: look for function_call blocks o 5 5
e@ Execute locally: run your function code
oe Send results back: as function_reault re rrr
e Repeat: until model returns text sheet So
ee : ee Sa
No history management
OQ Sevver tracks na meracton 10 SO Aromatic ma ; foot re
prenous meracton teak al
OCR text:
Boma et em ety Teetany Teta Cat Aeon oe Ca or i ne SC cee)
e ry
c a 4 o
mee
é Goo Ke
id Gg
Q Seacm Goze o type a URL $ BS @Amove
5 : . : , 4,
' - a
OCR text:
Cursor
Fle
Selection
Vew
COGCL
dAor813:25
SLLmX
ageetsskis)
SOLLmd
GeiasInteractons AI Saill
Curre Nodes （Ose Tese）
woratep
nan.Py
-flavitftlncedfct
n-3.1-gr-evie：tesclereanng,coding,reearch
(ss-lock)
geaini-3.1-flash-ite-srevjev1cost-efficient,fastest perfarance forhigh-frep
p-3---ee:5/3ktes,igrataeanitng
oei-2.5-re tescmlexresaning.coding.resarch
gemini-3.1-fLash-1nage-greviev65/32k toens,pegeerationandediting
geisi-2.5-flsh:tokens,fast,bolanceperfonnce,muttiaoda
Current Agents （Use These）
e-resrc-r-pei-12-225：DgResrchagt
[NG]
Carret SoKs Cse These）
5cript/Tyecripteogle/gal1.33.neInstalL@google/gea1
LICAUTTON)
egogie/pnerative-al（25）srydeprecated
Opt
Pors
ToolCatls!
（-0x9)
Soccess Rater
Perferm
otLTIel
2h3225s
entActivel
APT Tiae)
Tool Tinel
9s(0.)
hlscmld-ec[13:25:21][-/grects/ale
-itteractioes
OCR text:
, _ l
_
ee Tr
nn a a
an
OCR text:
README Apache-2.0 license Security
Skill for developing Gemini-powered apps. Provides the best practices for building apps that use the Gemini API.
Skill for developing Gemini-powered apps on Google Cloud Vertex AI using the Gen AI SDK. Covers tools, multimodal generation, caching, and batch prediction.
Skill for building real-time, bidirectional streaming apps with the Gemini Live API. Covers WebSocket-based audio/video/text streaming, voice activity detection, native audio features, function calling, and session management.
Skill for building apps with the Gemini Interactions API. Covers text generation, multi-turn chat, streaming, function calling, structured output, image generation, Deep Research agents, deprecated model guardrails, and both Python and TypeScript SDKs.
Installation
You can browse and install skills using either the Vercel skills CLI or the Context7 skills CLI.
Using Vercel skills CLI
# Interactively browse and install skills.
npx skills add google-gemini/gemini-skills --list
# Install a specific skill (e.g., gemini-api-dev).
npx skills add google-gemini/gemini-skills --skill gemini-api-dev --global
Using Context7 skills CLI
# Interactively browse and install skills.
npx ctx7 skills install google-gemini/gemini-skills
# Install a specific skill (e.g., vertex-ai-api-dev).
npx ctx7 skills install google-gemini/gemini-skills --skill vertex-ai-api-dev
Disclaimer
This is not an officially supported Google product. This project is not eligible for the Google Open Source Software Vulnerability Rewards Program.
OCR text:
Pe et er ce ed © > 6 Ok OF @ . EE oaetetcry
ee sie-workshop — main. py Open Agent Manager 03 (0 2 EB ea @
@ manpy x — 1) Walkthrough t2 Task @ impementation Plan > C €& > + Creating Agent Class Structure 7S x
wet og Me rarity
39 Class Agent: Next wo need to add a read_filo and write_tike tool.
69 def runiself, input_text: str | listh -> ste: . Create a basic python anplementation and atso the json
Ba for output in interaction.outputs: schema defition and map for the key: function, schema *
a9 Af output. type e2 “function_cali":
92 if outout.nane in self. tools:
1 TESMEL = IMC LAGE LULSa RL eeUUK pul el yume Ara sora CQ selupen
% except Exception a1 ¢:
47 result « f"Error executing tool: (strie)}™ . Analyzed @ agent.py: ”
a8 else: .
a result © (Error: Tool "{output.name}’ not found.” Vriagnt tet os
7 180
181 Prantif*OEBUG: Tool Result -» {result}”) Detining New Tools
1e? PL ALLALENO BRE 4
183 # Fospat the result as 9 function result biock SolUtion/agentapy tag .de re THe task st
18s tool_results. appendi { er , ee TINE TF
HE BOE Ot snd we ce ws
res type": “Sunction_result", ee eee eee ees
326 “calt_ie": output. id, workshop/main.py om sore por the Sy th or
187 "heme": output.nase, hp emestabars Ire GION ache ydeGstocs ane
1e8 Tresult": ste(resulet . Nap So Cee reel eR hte paesg to dy fob and
129 » neg
io
in a tf the eotel requested toot calls, we eust send the results back and con Fates tide -
12 if tool_results:
-
m return self. runitool_results) Edtec implementation Plan
ws # Return the teat frow the last output Gloca tususily the final response 6 I nave created an implementation plan to ada the
36 return interaction.outputsi-1}.text oxy ;
iw read_fite and write _file tools tothe Agent . Ths
118 def masini): includes ine Python rplementations, JSON schemas, and
# Instantiate the agent updaing the agent's mteracton loop ta handle tool calls.
agent = Agent() e e Be ray Co acces
@ First turn wo
erint("-—- Turn 1 --—")
question! = “Hi Reet: = 7 oy Parag 0 devin Veta os
ee ae | CRP RMT LE To dct a ot A a ee ea oe SCT a)
OCR text:
P|
le | oe?
" “3
Pie
oar
oy
ore
os ns fa
“as
! oe
! Tut
i Fe
| are
| Cts
aie
. Ee
rs
B aF
;
a
ov
' om a . .
| : li |
; y,
_
s oy
OCR text:
Gemini 3.1 Flash Live
OCR text:
nenu aistudio.go Lcom/a0ps/32166532-e57d-4375
[Public] Live Jukebox Remix o Device
Live uhebox
POWEREOBYGEMINILIVESLYRIR
SESSIONLOGS OTRROKS
STUOIDR SESSION
No tracksgeneratedyet Wish fora song to begin.
WAKNGLPTHEOLL
TALKTOOU
Text the DJyoutrequnst.
OCR text:
Live Jukebox
OCR text:
°
Live API
The Live API enables low-latency, real-time voice and vision interactions with Gemini.
ese Text, audio, video
“ee Text, audio +
App Connect via WebSocket Live API
OCR text:
a 5 _ oe
ro) A Ey . :
Google Al Studio - = Playground =
BS Playground >
% eula , :
i A
Ge Dasnboara > ld a A,
7 i
1% Documentation 2 \
id a EB
Voice
CoO/00e —— @ |
Media resolution
What do you think of my outta?
hse Thinking level
v
Ph 000/000 —=—
Well now, you're lookin’ very smart so you are!’ That green packet sutts you wed, | must say. A grand casual look. Sear ce Crepe os
Were you tunkin' of headin’ out somewhere?
OD) eee Atereesey many Mane MeAtaRes by oui ERR CARE
Disconnect renee A
& Search xe
Streamistve 8 Function caikng e
C Wat's new
exe Get APL key Suart typ ea prow pe Automatic Function Response e
® Sersngs
+ 8 O Grounding wth Google Search
@ thorsten schaetterg "a ®
OCR text:
wooabooon
D
GoogleAI Studlio
Playground
Choose what to share with aistudio.google.com
Run setings
<>Getcode
s oA o
Playground
Chror
Window
Bolld
System instructions
CoDathboard
Documentation
Yeah.
Voice
Ah,hello there! Whatrat
Mediaresolution
Thinking level
Q.Search
Stream is live
Function calling
0what'snew
Get APIkoy
Settings
GGrounding with Google SearchX
Grounding with
Source:G
thorsten.schaeff@g
Google Search
Google Search
OCR text:
4
iy Peck G9
tT Lk peat Le ge 7
: | ¥ Cr A ade a j 4 i at awe) an] a Eco f.. . x q 1
ar pe ees op <a Pe “) ee oe er
..- Smilin ae
toa ee a i é i a us i
a a rs ee.
‘oo 5 ig an ne a en re { . 4, &
a : ee UP eee es . 7 pan 3
a, es BO ps Fa een mer _—
ae oe ne ; “as 7 7 rer,
- aaa ie on r yy oe aa) es yt!
< a Ps Ld ‘ a a * il ae
a a i a an aN H
nn ce : _— Se ors ;
i
a 7 aie at ono oe Oo
, aan | ie ware
_ id —
,
cu |: cae ome
OCR text:
ged mi
Be — re pi
oe aes iy ot
eee Pavel so anne ri
‘ea te ee -~ :
a % ®
oe 4 7 Big ;
Pe ae! ieee at ce
an . roa
oe al Brot 399
‘ty ;
rf A
; :
FAT Sawn
c I ao os rn ry “oy r s aad avr "@ a + a oe
— i aa *. 2° : .
a a Pe _. eee eo:
Geman
Slide-Derived Subjects To Review
Subject extraction uses video title, related session titles/descriptions, transcript context, and OCR text when available. OCR is best-effort and should be reviewed against the embedded slide images.