tzolov.bsky.social
Spring AI lead | Spring Framework team at Broadcom
34 posts
273 followers
118 following
Regular Contributor
Active Commenter
comment in response to
post
4️⃣ Beyond the Hype: Building AI Agents with Spring /
@asaikali.bsky.social & @tzolov.bsky.social
Get a clear and practical introduction to building AI agents with Spring AI.
🔗 2025.springio.net/workshops/be...
comment in response to
post
Love it! Also like the Bulgarian folk polyphony touch ;)
comment in response to
post
To test streaming tools, use Spring AI 1.0.0-SNAPSHOT, as version 1.0.0-M4 predates Ollama's streaming support. Update your project's pom.xml to:
- Change the dependency to 1.0.0-SNAPSHOT
- Add the Spring Snapshot Maven repository (details at docs.spring.io/spring-ai/re...)
Hope this helps
comment in response to
post
Hi Freddy,
What version of Spring AI are you using?
Ollama implemented Streaming tool calls just 2 weeks ago.
So we provided integration for it: github.com/spring-proje...
But you need to use the current 1.0.0-SNAPSHOT or wait for M5
comment in response to
post
Yes, mcp-core and mcp-spring are designed to enable MCP server implementation in Java. We're currently finalizing the remote transport components (HTTP/POST and SSE), with server MCP development upcoming. Most necessary components are already in place.
github.com/spring-proje...
comment in response to
post
github.com/tzolov/sprin...
comment in response to
post
Here is the video: youtu.be/LcKHqZ8oi9w?...
The interesting part, e.g. the Flight Booking Assistant demo, starts around ~ 9min
comment in response to
post
Douglas Adams is remarkable and inspirational ... The Long Dark Tea-Time of the Soul - the limbo time on a Sunday, after afternoon but before evening, as the weekend had finished but the week had not yet begun ...
comment in response to
post
Yep, a classical guitar arrangement of paranoid android by RadioHead :)
comment in response to
post
Didn't know about Cline. Very nice.
comment in response to
post
Recently someone mentioned the "Language Server for Java by Apache NetBeans" as an alternative to the mainstream ReadHat/Eclipse one, but I've never tried it.
comment in response to
post
VS Code's been my go-to Java IDE since 2022. Configuration and shortcuts were a learning curve. Getting to know settings.json might still be required. The "IntelliJ IDEA Keybindings" extension can help to transition the shortcuts.
comment in response to
post
One last though about the APIs support.
Ollama support the OpenAI API as as well. That means you can use the Open AI client to talk to Ollama.
Spring AI allows you to access Ollama over OpenAI API like this: docs.spring.io/spring-ai/re...
comment in response to
post
Spring AI's low-level Ollama API (class is called OllamaApi) still provides access to the /api/generate endpoint. But it doesn't make sense to use it for high-level APIs as the `/api/chat` handles all cases.
comment in response to
post
Historically, the /api/generate was the first Ollama API for chat completions, but it was non sufficient for conversation cases. The `/api/generate` had a (deprecated now) `context` field to try to handle this but, obviously this wasn't enough and Ollama provided /api/chat as complete solution.
comment in response to
post
Hi Freddy, as discussed there github.com/ollama/ollam...
the only difference between `api/chat` and `api/generate` is that the former can, OPTIONALLY, add chat history and the later can't. If you don't add the history explicitly, then only one message (one-shot) is sent and both are equivalent.
comment in response to
post
You can use HuggingFace models via Ollama: docs.spring.io/spring-ai/re...
Ollama supports multimodality: docs.spring.io/spring-ai/re...
If you are interested in extending the @spring-ai.bsky.social HuggingFace integration, you are welcome. Haven't seen much community traction about it.
comment in response to
post
Next to Vertex Gemini, the Anthropic Claude is another AI Model supporting PDF multimodal input:
- Anthropic Claude - docs.spring.io/spring-ai/re...
- Vertex Gemini - docs.spring.io/spring-ai/re...
comment in response to
post
And so hard
comment in response to
post
The comparison table was also missing the new Amazon Bedrock Converse API integration, which introduces extensive features like Tools and Modality support. Amazon recommends this as their primary Chat/Message API moving forward:
comment in response to
post
Table is updated, listing the Gemini PDF modality support.
The IT is here: shorturl.at/Rwraj
It reminds me that we have to enable the Anthropic PDF modality support as well.
comment in response to
post
Nice article and historically accurate: “One of the earliest examples of this was the Atanasoff-Berry “computer” (ABC) “