v3.4.1.1

New Features:

• MessageCenter
• EventEmitter

Full Changelog: v3.4.1.0...v3.4.1.1

v3.4.1.0

New Feature: new utils Stage and Tunnel to help manage coroutine and threads

Update: rewrite ResponseGenerator using Stage and Tunnel

Bug Fix: fixed a bug that will cause agent can not inhert debug setti…

Bug Fix: change param name for new version of ddgs

Full Changelog: v3.4.0.5...v3.4.1.0

v3.4.0.5

What's Changed

• Fixed broken link in Readme.md by @Shreyas0410 in #178
• refactor: Refactored Status.py by @ghimirebibek in #179
• update by @Maplemx in #182
• v3.4.0.5 by @Maplemx in #183

New Contributors

• @Shreyas0410 made their first contribution in #178
• @ghimirebibek made their first contribution in #179

Full Changelog: v3.4.0.4...v3.4.0.5

v3.4.0.4

Instant Mode (Former Realtime Mode)

1. Rename "Realtime" component to "Instant" to avoid improper associate to OpenAI Realtime;
2. Optimize key indexes combination expression handler and add "&" symbol support;
3. Remove ".$complete" mark for string items and add ".$delta" mark to make string items behave the same as other type items;

Response Generator

1. Rewrite ResponseGenerator component to provide 4 agentic request response generators, developers can use these alias down below to get generator type response instead of agent.start():

   • agent.get_generator(): classic response generator including events start, delta, done
   • agent.get_complete_generator(): response generator including all events during request
   • agent.get_instant_generator(): response generator including response data in Instant Mode
   • agent.get_instant_keys_generator(): response generator including response data when specific key indexes combination occur

   Read this document for more details

Full Changelog: v3.4.0.3...v3.4.0.4

v3.4.0.3

New Feature:

• ResponseGenerator: a better way to get streaming response generator from Agently AgenticRequest, try these codes down below to feel the better develop experience provided by Agently Realtime x ResponseGenerator:

generator = (
    agent
        .input("Generator 10 sentences")
        .output({
            "sentences": ([("str", )]),
        })
        .get_realtime_generator()
)

for item in generator:
    print(item["key"], item["delta"])

We also provide .get_complete_generator() for you to get all events (including response:delta_origin, response:done_origin with detail information) or .get_generator() to get classic event response:start, response:delta and response:done.

Updates:

• Realtime mode will be automatically turn on when developers try to add an realtime event listener or use realtime generator, don't worry about forget to add .use_realtime() to your chain syntax anymore.
• use settings to control use_realtime status to make this status better to be used in other agent components.

Full Changelog: v3.4.0.2...v3.4.0.3

v3.4.0.2

What's Changed

• Update Realtime by @Maplemx in #172

Full Changelog: v3.4.0.1...v3.4.0.2

v3.4.0.1

What's Changed

• Add delta to realtime response data by @Maplemx in #171

Full Changelog: v3.4.0.0...v3.4.0.1

v3.4.0.0

(Will update soon)

What's Changed

• Update README.md by @moyueheng in #167
• keep up by @Maplemx in #169
• Dev by @Maplemx in #170

New Contributors

• @moyueheng made their first contribution in #167

Full Changelog: v3.3.4.8...v3.4.0.0

v3.3.4.8

Update

1. [AppConnector] Support Gradio additional inputs.
2. [DataGenerator] Add new methods .future(), .join() to support pre-define generator handler before .add() is called.

v3.3.4.7

Yeah... I'm sorry that I didn't update release document here in time and as you know... just like a small snowball rolling down from the top of a slope, it growth bigger and bigger, and it gets harder and harder for me to update this release document...

In fact, from v3.2.2.3 to v.3.3.4.7 we make so much progress such as:

• We improved Agently Workflow to make it both powerful and easy to use, read Agently Workflow Official Document to explore more.
• We add supports of a lot of models, read Agently AnyModel Official Document
• We add a new feature named AppConnector that will help developers to create an application with web UI really quick, and developers can switch application UI framework between gradio, streamlit and shell without changing any application logic code. In version 3.3.4.x we will continue to improve this feature.
• And of course we fixed a lot of bugs and did a lot of optimizing work.

As a really small team with just one full time developer and one part time developer, maybe we did not so good as those team with enough human resource and money support, but we're still keep fighting and keep going because we believe Agently AI application development framework can really help developers to create wonderful LLM based applications faster and easier.

We'll do better! Fighting!