{ "@context": "https://schema.org", "@id": "https://aggr.aymericbeaumet.com/items/philschmid-de/2026-02-02-getting-started-with-gemini-deep-research-api/#webpage", "@type": [ "WebPage", "ArchiveComponent" ], "articleBody": "This guide demonstrates how to use the new Gemini Deep Research agent via the [Interactions API](https://ai.google.dev/gemini-api/docs/interactions?ua=chat) to perform complex research tasks, generate images based on the findings, and translate the results.\n\n**Preview:** The Gemini Deep Research Agent is currently in preview. The Deep Research agent is exclusively available using the [Interactions API](https://ai.google.dev/gemini-api/docs/interactions). You cannot access it through `generate_content`.\n\nPython\n\n```python\n!uv pip install google-genai --upgrade\n```\n\nPython\n\n```python\nimport time\nfrom google import genai\nfrom IPython.display import display, Markdown\n \nclient = genai.Client()\n```\n\n## 1\\. Basic Research Task (Polling)\n\nThe following example shows how to start a research task in the background and poll for results. This is the standard way to interact with the Deep Research agent.\n\nPython\n\n```python\nprompt = \"Research the history of Google TPUs.\"\n \ninteraction = client.interactions.create(\n agent='deep-research-pro-preview-12-2025',\n input=prompt,\n background=True\n)\n \nprint(f\"Research started: {interaction.id}\")\n# Poll for results\nwhile True:\n interaction = client.interactions.get(interaction.id)\n if interaction.status == \"completed\":\n print(f\"Research completed: {interaction.id}\")\n break\n elif interaction.status == \"failed\":\n print(f\"Research failed: {interaction.error}\")\n break\n \n print(\".\", end=\"\", flush=True)\n time.sleep(10)\n \n \n \ndisplay(Markdown(interaction.outputs[-1].text))\n```\n\n```plaintext\n# The Evolution of Google Tensor Processing Units: A Historical and Architectural Analysis\n \n### Key Points\n* **Origins in Necessity:** The Tensor Processing Unit (TPU) project was initiated in 2013 when Google engineers realized that existing CPU and GPU infrastructure could not sustain the computational demands of ubiquitous voice search and deep learning applications without doubling datacenter footprints.\n* **Architectural Paradigm Shift:** Unlike the Von Neumann architecture of CPUs, TPUs utilize a **systolic array** architecture. This design mimics the flow of blood through the heart, allowing data to flow through thousands of arithmetic logic units (ALUs) without accessing memory for intermediate results, drastically increasing efficiency for matrix multiplications.\n...\n```\n\nDeep Research supports streaming to receive real-time updates on the research progress (`thinking_summaries`). You can steer the agent's output by providing specific promting instructions.\n\nPython\n\n```python\n# Define a complex prompt with formatting instructions\nprompt = \"\"\"\nResearch the history and future developments of the Google TPUs.\n \nFocus on:\n1. Key technological breakthroughs in the last 12 months.\n2. Major competitors and their projected timelines for mass production.\n3. Future developments and plans for 2026 and beyond\n \nFormat the output:\n- as a strategic briefing document for an developer educating themselves about the Google TPU landscape.\n- Include a table comparing the top 3 leading companies.\n- Use short concise sentence and bullet points.\n\"\"\"\n \n \nstream = client.interactions.create(\n agent=\"deep-research-pro-preview-12-2025\",\n input=prompt,\n background=True,\n stream=True,\n agent_config={\n \"type\": \"deep-research\",\n \"thinking_summaries\": \"auto\"\n }\n)\n \ninteraction_id = None\nlast_event_id = None\nreport = \"\"\n \nfor chunk in stream:\n if chunk.event_type == \"interaction.start\":\n interaction_id = chunk.interaction.id\n print(f\"Interaction started: {interaction_id}\")\n \n if chunk.event_id:\n last_event_id = chunk.event_id\n \n if chunk.event_type == \"content.delta\":\n if chunk.delta.type == \"text\":\n report += chunk.delta.text\n elif chunk.delta.type == \"thought_summary\":\n print(f\"{chunk.delta.content.text}\\n\", flush=True)\n \n elif chunk.event_type == \"interaction.complete\":\n print(chunk)\n print(\"\\nResearch Complete\")\n \n \nprev_ia = client.interactions.get(interaction_id)\nprint(prev_ia)\n \ndisplay(Markdown(report))\n \n```\n\nFull Report: [https://gist.github.com/philschmid/68c5afacfd3a3555ca834ba27415ba88](https://gist.github.com/philschmid/68c5afacfd3a3555ca834ba27415ba88)\n\n## 3\\. Combine Deep Research Interaciton with Model Interactions\n\nThe Interactions API supports stateful interactions, which allows you to continue the conversation after the agent returns the final report by using the `previous_interaction_id`.\n\nYou can continue the conversation after the agent returns the final report using a model to visualize the results using Nano Banana Pro or translate the results using Gemini Flash.\n\nPython\n\n```python\nimport base64\nfrom IPython.display import Image\n \nprint(\"Generating image...\")\nimage_interaction = client.interactions.create(\n model=\"gemini-3-pro-image-preview\",\n input=\"Visluaize the report ino a timeline slide\",\n previous_interaction_id=interaction_id,\n)\n \nimage_data = [data for data in image_interaction.outputs if data.type == \"image\"]\n \nfor output in image_data:\n image_data = base64.b64decode(output.data)\n display(Image(data=image_data))\n```\n\n![timeline](https://www.philschmid.de/static/blog/gemini-deep-research-getting-started/timeline.jpg)\n\nPython\n\n```python\nprint(\"Translating...\")\ntranslate_interaction = client.interactions.create(\n model=\"gemini-3-flash-preview\",\n input=\"Translate the report into simple German.\",\n previous_interaction_id=interaction_id,\n)\n \ndisplay(Markdown(translate_interaction.outputs[-1].text))\n```\n\n```plaintext\n# Strategischer Bericht: Die Google TPU-Landschaft\n \n## Zusammenfassung\nGoogle beschleunigt die Entwicklung eigener KI-Chips (TPUs), um NVIDIA herauszufordern. Die letzten 12 Monate brachten mit **Trillium (v6)** und **Ironwood (v7)** große Fortschritte. Für Entwickler wird es einfacher: Dank neuer Software-Tools laufen PyTorch-Modelle jetzt fast ohne Änderungen auf TPUs. Bis 2027 will Google über 5 Millionen TPUs im Einsatz haben.\n... \n```\n\n* * *\n\n**This API is in Beta, and we want your feedback!** We're actively listening to developers to shape the future of this API. What features would help your agent workflows? What pain points are you experiencing? Please let me know on [Twitter](https://twitter.com/_philschmid) or [LinkedIn](https://www.linkedin.com/in/philipp-schmid-a6a2bb196/).\n", "authors": [], "captured_at": "2026-09-27T19:29:15.293681927+00:00", "category": "programming", "content_html": "

This guide demonstrates how to use the new Gemini Deep Research agent via the Interactions API to perform complex research tasks, generate images based on the findings, and translate the results.

\n

Preview: The Gemini Deep Research Agent is currently in preview. The Deep Research agent is exclusively available using the Interactions API. You cannot access it through generate_content.

\n

Python

\n
!uv pip install google-genai --upgrade\n
\n

Python

\n
import time\nfrom google import genai\nfrom IPython.display import display, Markdown\n \nclient = genai.Client()\n
\n

1. Basic Research Task (Polling)

\n

The following example shows how to start a research task in the background and poll for results. This is the standard way to interact with the Deep Research agent.

\n

Python

\n
prompt = "Research the history of Google TPUs."\n \ninteraction = client.interactions.create(\n    agent='deep-research-pro-preview-12-2025',\n    input=prompt,\n    background=True\n)\n \nprint(f"Research started: {interaction.id}")\n# Poll for results\nwhile True:\n    interaction = client.interactions.get(interaction.id)\n    if interaction.status == "completed":\n        print(f"Research completed: {interaction.id}")\n        break\n    elif interaction.status == "failed":\n        print(f"Research failed: {interaction.error}")\n        break\n    \n    print(".", end="", flush=True)\n    time.sleep(10)\n \n \n \ndisplay(Markdown(interaction.outputs[-1].text))\n
\n
# The Evolution of Google Tensor Processing Units: A Historical and Architectural Analysis\n \n### Key Points\n*   **Origins in Necessity:** The Tensor Processing Unit (TPU) project was initiated in 2013 when Google engineers realized that existing CPU and GPU infrastructure could not sustain the computational demands of ubiquitous voice search and deep learning applications without doubling datacenter footprints.\n*   **Architectural Paradigm Shift:** Unlike the Von Neumann architecture of CPUs, TPUs utilize a **systolic array** architecture. This design mimics the flow of blood through the heart, allowing data to flow through thousands of arithmetic logic units (ALUs) without accessing memory for intermediate results, drastically increasing efficiency for matrix multiplications.\n...\n
\n

Deep Research supports streaming to receive real-time updates on the research progress (thinking_summaries). You can steer the agent's output by providing specific promting instructions.

\n

Python

\n
# Define a complex prompt with formatting instructions\nprompt = """\nResearch the history and future developments of the Google TPUs.\n \nFocus on:\n1.  Key technological breakthroughs in the last 12 months.\n2.  Major competitors and their projected timelines for mass production.\n3.  Future developments and plans for 2026 and beyond\n \nFormat the output:\n- as a strategic briefing document for an developer educating themselves about the Google TPU landscape.\n- Include a table comparing the top 3 leading companies.\n- Use short concise sentence and bullet points.\n"""\n \n \nstream = client.interactions.create(\n    agent="deep-research-pro-preview-12-2025",\n    input=prompt,\n    background=True,\n    stream=True,\n    agent_config={\n        "type": "deep-research",\n        "thinking_summaries": "auto"\n    }\n)\n \ninteraction_id = None\nlast_event_id = None\nreport = ""\n \nfor chunk in stream:\n    if chunk.event_type == "interaction.start":\n        interaction_id = chunk.interaction.id\n        print(f"Interaction started: {interaction_id}")\n \n    if chunk.event_id:\n        last_event_id = chunk.event_id\n \n    if chunk.event_type == "content.delta":\n        if chunk.delta.type == "text":\n            report += chunk.delta.text\n        elif chunk.delta.type == "thought_summary":\n            print(f"{chunk.delta.content.text}\\n", flush=True)\n \n    elif chunk.event_type == "interaction.complete":\n        print(chunk)\n        print("\\nResearch Complete")\n        \n \nprev_ia = client.interactions.get(interaction_id)\nprint(prev_ia)\n \ndisplay(Markdown(report))\n \n
\n

Full Report: https://gist.github.com/philschmid/68c5afacfd3a3555ca834ba27415ba88

\n

3. Combine Deep Research Interaciton with Model Interactions

\n

The Interactions API supports stateful interactions, which allows you to continue the conversation after the agent returns the final report by using the previous_interaction_id.

\n

You can continue the conversation after the agent returns the final report using a model to visualize the results using Nano Banana Pro or translate the results using Gemini Flash.

\n

Python

\n
import base64\nfrom IPython.display import Image\n \nprint("Generating image...")\nimage_interaction = client.interactions.create(\n    model="gemini-3-pro-image-preview",\n    input="Visluaize the report ino a timeline slide",\n    previous_interaction_id=interaction_id,\n)\n \nimage_data = [data for data in image_interaction.outputs if data.type == "image"]\n \nfor output in image_data:\n    image_data = base64.b64decode(output.data)\n    display(Image(data=image_data))\n
\n

\"timeline\"

\n

Python

\n
print("Translating...")\ntranslate_interaction = client.interactions.create(\n    model="gemini-3-flash-preview",\n    input="Translate the report into simple German.",\n    previous_interaction_id=interaction_id,\n)\n \ndisplay(Markdown(translate_interaction.outputs[-1].text))\n
\n
# Strategischer Bericht: Die Google TPU-Landschaft\n \n## Zusammenfassung\nGoogle beschleunigt die Entwicklung eigener KI-Chips (TPUs), um NVIDIA herauszufordern. Die letzten 12 Monate brachten mit **Trillium (v6)** und **Ironwood (v7)** große Fortschritte. Für Entwickler wird es einfacher: Dank neuer Software-Tools laufen PyTorch-Modelle jetzt fast ohne Änderungen auf TPUs. Bis 2027 will Google über 5 Millionen TPUs im Einsatz haben.\n... \n
\n
\n

This API is in Beta, and we want your feedback! We're actively listening to developers to shape the future of this API. What features would help your agent workflows? What pain points are you experiencing? Please let me know on Twitter or LinkedIn.

\n", "content_markdown": "This guide demonstrates how to use the new Gemini Deep Research agent via the [Interactions API](https://ai.google.dev/gemini-api/docs/interactions?ua=chat) to perform complex research tasks, generate images based on the findings, and translate the results.\n\n**Preview:** The Gemini Deep Research Agent is currently in preview. The Deep Research agent is exclusively available using the [Interactions API](https://ai.google.dev/gemini-api/docs/interactions). You cannot access it through `generate_content`.\n\nPython\n\n```python\n!uv pip install google-genai --upgrade\n```\n\nPython\n\n```python\nimport time\nfrom google import genai\nfrom IPython.display import display, Markdown\n \nclient = genai.Client()\n```\n\n## 1\\. Basic Research Task (Polling)\n\nThe following example shows how to start a research task in the background and poll for results. This is the standard way to interact with the Deep Research agent.\n\nPython\n\n```python\nprompt = \"Research the history of Google TPUs.\"\n \ninteraction = client.interactions.create(\n agent='deep-research-pro-preview-12-2025',\n input=prompt,\n background=True\n)\n \nprint(f\"Research started: {interaction.id}\")\n# Poll for results\nwhile True:\n interaction = client.interactions.get(interaction.id)\n if interaction.status == \"completed\":\n print(f\"Research completed: {interaction.id}\")\n break\n elif interaction.status == \"failed\":\n print(f\"Research failed: {interaction.error}\")\n break\n \n print(\".\", end=\"\", flush=True)\n time.sleep(10)\n \n \n \ndisplay(Markdown(interaction.outputs[-1].text))\n```\n\n```plaintext\n# The Evolution of Google Tensor Processing Units: A Historical and Architectural Analysis\n \n### Key Points\n* **Origins in Necessity:** The Tensor Processing Unit (TPU) project was initiated in 2013 when Google engineers realized that existing CPU and GPU infrastructure could not sustain the computational demands of ubiquitous voice search and deep learning applications without doubling datacenter footprints.\n* **Architectural Paradigm Shift:** Unlike the Von Neumann architecture of CPUs, TPUs utilize a **systolic array** architecture. This design mimics the flow of blood through the heart, allowing data to flow through thousands of arithmetic logic units (ALUs) without accessing memory for intermediate results, drastically increasing efficiency for matrix multiplications.\n...\n```\n\nDeep Research supports streaming to receive real-time updates on the research progress (`thinking_summaries`). You can steer the agent's output by providing specific promting instructions.\n\nPython\n\n```python\n# Define a complex prompt with formatting instructions\nprompt = \"\"\"\nResearch the history and future developments of the Google TPUs.\n \nFocus on:\n1. Key technological breakthroughs in the last 12 months.\n2. Major competitors and their projected timelines for mass production.\n3. Future developments and plans for 2026 and beyond\n \nFormat the output:\n- as a strategic briefing document for an developer educating themselves about the Google TPU landscape.\n- Include a table comparing the top 3 leading companies.\n- Use short concise sentence and bullet points.\n\"\"\"\n \n \nstream = client.interactions.create(\n agent=\"deep-research-pro-preview-12-2025\",\n input=prompt,\n background=True,\n stream=True,\n agent_config={\n \"type\": \"deep-research\",\n \"thinking_summaries\": \"auto\"\n }\n)\n \ninteraction_id = None\nlast_event_id = None\nreport = \"\"\n \nfor chunk in stream:\n if chunk.event_type == \"interaction.start\":\n interaction_id = chunk.interaction.id\n print(f\"Interaction started: {interaction_id}\")\n \n if chunk.event_id:\n last_event_id = chunk.event_id\n \n if chunk.event_type == \"content.delta\":\n if chunk.delta.type == \"text\":\n report += chunk.delta.text\n elif chunk.delta.type == \"thought_summary\":\n print(f\"{chunk.delta.content.text}\\n\", flush=True)\n \n elif chunk.event_type == \"interaction.complete\":\n print(chunk)\n print(\"\\nResearch Complete\")\n \n \nprev_ia = client.interactions.get(interaction_id)\nprint(prev_ia)\n \ndisplay(Markdown(report))\n \n```\n\nFull Report: [https://gist.github.com/philschmid/68c5afacfd3a3555ca834ba27415ba88](https://gist.github.com/philschmid/68c5afacfd3a3555ca834ba27415ba88)\n\n## 3\\. Combine Deep Research Interaciton with Model Interactions\n\nThe Interactions API supports stateful interactions, which allows you to continue the conversation after the agent returns the final report by using the `previous_interaction_id`.\n\nYou can continue the conversation after the agent returns the final report using a model to visualize the results using Nano Banana Pro or translate the results using Gemini Flash.\n\nPython\n\n```python\nimport base64\nfrom IPython.display import Image\n \nprint(\"Generating image...\")\nimage_interaction = client.interactions.create(\n model=\"gemini-3-pro-image-preview\",\n input=\"Visluaize the report ino a timeline slide\",\n previous_interaction_id=interaction_id,\n)\n \nimage_data = [data for data in image_interaction.outputs if data.type == \"image\"]\n \nfor output in image_data:\n image_data = base64.b64decode(output.data)\n display(Image(data=image_data))\n```\n\n![timeline](https://www.philschmid.de/static/blog/gemini-deep-research-getting-started/timeline.jpg)\n\nPython\n\n```python\nprint(\"Translating...\")\ntranslate_interaction = client.interactions.create(\n model=\"gemini-3-flash-preview\",\n input=\"Translate the report into simple German.\",\n previous_interaction_id=interaction_id,\n)\n \ndisplay(Markdown(translate_interaction.outputs[-1].text))\n```\n\n```plaintext\n# Strategischer Bericht: Die Google TPU-Landschaft\n \n## Zusammenfassung\nGoogle beschleunigt die Entwicklung eigener KI-Chips (TPUs), um NVIDIA herauszufordern. Die letzten 12 Monate brachten mit **Trillium (v6)** und **Ironwood (v7)** große Fortschritte. Für Entwickler wird es einfacher: Dank neuer Software-Tools laufen PyTorch-Modelle jetzt fast ohne Änderungen auf TPUs. Bis 2027 will Google über 5 Millionen TPUs im Einsatz haben.\n... \n```\n\n* * *\n\n**This API is in Beta, and we want your feedback!** We're actively listening to developers to shape the future of this API. What features would help your agent workflows? What pain points are you experiencing? Please let me know on [Twitter](https://twitter.com/_philschmid) or [LinkedIn](https://www.linkedin.com/in/philipp-schmid-a6a2bb196/).\n", "dateCreated": "2026-09-27T19:29:15.293681927+00:00", "date_created": "2026-02-02T00:00:00+00:00", "date_modified": "2026-02-02T00:00:00+00:00", "date_published": "2026-02-02T00:00:00+00:00", "description": "Learn how to use the new Gemini Deep Research agent via the Interactions API to perform complex research tasks, generate images based on the findings, and translate the results.", "external_url": "https://www.philschmid.de/gemini-deep-research-getting-started", "git": { "commit": "ebb0941c5c06a2ab2377342d607535262a3e969e", "path": "items/philschmid-de/2026/02/2026-02-02-getting-started-with-gemini-deep-research-api.md", "permalink": "https://github.com/aymericbeaumet/aggr-instance/blob/ebb0941c5c06a2ab2377342d607535262a3e969e/items/philschmid-de/2026/02/2026-02-02-getting-started-with-gemini-deep-research-api.md" }, "id": "urn:aggr:item:3d50126a93f2abba7c98cab50094f5c12fb47ea8", "isBasedOn": { "@id": "https://www.philschmid.de/gemini-deep-research-getting-started" }, "mainEntity": { "@id": "https://www.philschmid.de/gemini-deep-research-getting-started", "@type": "CreativeWork", "archivedAt": { "@id": "https://aggr.aymericbeaumet.com/items/philschmid-de/2026-02-02-getting-started-with-gemini-deep-research-api/#webpage" }, "dateModified": "2026-02-02T00:00:00+00:00", "datePublished": "2026-02-02T00:00:00+00:00", "headline": "Getting Started with Gemini Deep Research API", "timeRequired": "PT4M", "url": "https://www.philschmid.de/gemini-deep-research-getting-started", "wordCount": 709 }, "name": "Archived snapshot: Getting Started with Gemini Deep Research API", "preview": null, "reading_minutes": 4, "replicated_at": null, "source": "philschmid-de", "source_name": "philschmid.de - RSS feed", "summary": "Learn how to use the new Gemini Deep Research agent via the Interactions API to perform complex research tasks, generate images based on the findings, and translate the results.", "tags": [], "temporalCoverage": "2026-09-27T19:29:15.293681927+00:00", "title": "Getting Started with Gemini Deep Research API", "url": "https://aggr.aymericbeaumet.com/items/philschmid-de/2026-02-02-getting-started-with-gemini-deep-research-api/", "word_count": 709 }