122 lines
6.6 KiB
Markdown
122 lines
6.6 KiB
Markdown
tags:: #todoist-task, [[PROJECTS/MOSAICO]]
|
|
date:: [[01-07-2025]] - 09:25
|
|
progress:: {{renderer :todomaster}}
|
|
|
|
- ### Tasks
|
|
-
|
|
- ### Notes
|
|
- Agenda: 
|
|
- Participants: 
|
|
- ### Todos
|
|
collapsed:: true
|
|
- {{query (and [[TODO]] [[Missione ROMA Mosaico]])}}
|
|
query-table:: true
|
|
query-properties:: [:block]
|
|
- **First day of meeting [[01-07-2025]]**
|
|
collapsed:: true
|
|
- ### Overview
|
|
collapsed:: true
|
|
- 
|
|
- ### WP1 presentation
|
|
collapsed:: true
|
|
- 
|
|
- 
|
|
- Where is the consensus agent?
|
|
- The shown workflow is an initial simplification. The idea is having more complex workflows where consensus agents are part of the workflow.
|
|
- 
|
|
- 
|
|
- This is the API of the server that is executing agents
|
|
- {:height 480, :width 951}
|
|
- TODO Look at Phoenix technology
|
|
- TODO WP2 and WP3 need to decide where to put the telemetry function
|
|
- 
|
|
- 
|
|
- 
|
|
- {:height 561, :width 951}
|
|
- 
|
|
- TODO Vedere LangFuse for observability.
|
|
- 
|
|
- You can see different calls / tools used
|
|
- Compare to Phoenix, the license is more friendly and you can use this completely locally.
|
|
- Opentelemetry is a communication protocol and LangFuse supports this protocol.
|
|
- 
|
|
- 
|
|
- TODO Da guardare A2A agent card per la tassonomia relativa al repository
|
|
- 
|
|
- 
|
|
- 
|
|
- TODO The agent discovery mechanism is very much related to that of MOSAICO. Have a look at them especially to see what is the taxonomy underpinning them
|
|
- 
|
|
- 
|
|
- 
|
|
- TODO This is an idea related to the WP2 taxonomy
|
|
- 
|
|
- ### WP2 presentation
|
|
collapsed:: true
|
|
- 
|
|
- TODO Developer Interface is not the right naming. THis needs to be changed
|
|
- 
|
|
- TODO LiteLLM can be seen as a possible alternative for the telemetry
|
|
- 
|
|
- TODO How do they implement guard rails?
|
|
- This is for security but also **reliability**
|
|
- TODO How scalable are they in terms of agents?
|
|
- TODO Do they support agents in different technologies?
|
|
- TODO What about parallel executions?
|
|
- TODO Improve the relation vs existing model repositories
|
|
- TODO Have a look at the different repositories shared by Antonio
|
|
- https://beeai.dev/agents
|
|
- https://www.agentlocker.ai/
|
|
- https://aiagentslist.com/
|
|
- [Taxonomy of AI Agent Skills - Agntcy](https://docs.agntcy.org/oasf/taxonomy/)
|
|
- TODO Look at Traces and Metrics (OpenTelemetry with Spring)
|
|
- Our application should be ready to gather telemetry
|
|
- It is related to LangFuse
|
|
- Traces to LangFuse
|
|
- Metrics go to Graphana
|
|
- OpenTelemetry does both
|
|
- We talk metrics but we should also talk about the evaluator
|
|
- TODO Check alternative technologies than MongoDB
|
|
- https://docs.spring.io/spring-ai/reference/api/vectordbs.html#_vectorstore_implementations
|
|
- TODO Link the taxonomy with the SEBOK
|
|
-
|
|
-
|
|
- ### WP3 Presentation
|
|
collapsed:: true
|
|
- 
|
|
- 
|
|
- Tipically used for developing agents in a rule-based manner
|
|
- How people thinks!
|
|
- 
|
|
- 
|
|
- How to make BDI more pervasive in MOSAICO
|
|
- (last point and see next slide)
|
|
- WE can select an agent with respect to "desires" typically writte in natural language
|
|
- Believes for selecting agents
|
|
- Agent will expose intentions while it's adoption (e.g. in vscode you can accept or not what agent is going to do)
|
|
- WP3 is kind of requiremnt for WP1 !?
|
|
- 
|
|
- 
|
|
- 
|
|
- 
|
|
- 
|
|
- This scenario comes from the interaction with Collins
|
|
-
|
|
- ### WP4 Presentation
|
|
- 
|
|
- 
|
|
- ... different profiles (from WP2)
|
|
- 
|
|
- 
|
|
- 
|
|
- {:height 532, :width 921}
|
|
- 
|
|
- 
|
|
- 
|
|
-
|
|
- 
|
|
- Attributes of the Agent class are related to the repository. For instance, the confidence of each agent should be available from the repo.
|
|
- The list of attributes of the Agent class should be refined by considering also who is going to value/compute them.
|
|
- 
|
|
- 
|
|
-  |