Berack/upo-app-agents
0
0
Fork 0
Code Issues Pull Requests 3 Actions Packages Projects Releases Wiki Activity
112 Commits 5 Branches 0 Tags
df65d36f4e3347fe50c6420614ca012c086a2141
Commit Graph

5 Commits

Author SHA1 Message Date
Berack96
df65d36f4e Load toolkit instructions from centralized instruction files 2025-10-31 00:08:12 +01:00
Simone Garau
a47ce46ea9 feat: Add comprehensive toolkit instructions and improve agent prompts 
- Create detailed markdown instructions for all toolkits:
  * market_instructions.md: 6 tools with VWAP aggregation, selection strategies
  * news_instructions.md: 4 tools with sentiment analysis guidelines
  * social_instructions.md: 2 tools with platform-specific notes
  * plan_memory_instructions.md: 4 tools with workflow patterns and best practices
  * symbols_instructions.md: 2 tools with symbol resolution and search

- Update all toolkit classes to load instructions from external .md files:
  * MarketAPIsTool, NewsAPIsTool, SocialAPIsTool
  * PlanMemoryTool (moved from agents to api/tools)
  * CryptoSymbolsTools

- Enhance agent prompts with detailed tool usage guidance:
  * team_market.md: Expanded tool descriptions, asset conversion, time ranges
  * team_news.md: Added query formulation, limit guidelines, error handling
  * team_social.md: Platform-specific notes, limit guidelines, content warnings
  * team_leader.md: Added CryptoSymbolsTools integration, detailed workflow steps

- Fix market data aggregation:
  * aggregate_multi_assets now calculates total volume (sum) instead of average
  * Implement currency filtering (prefer USD when currencies differ)
  * Refactor aggregate_multi_assets to call aggregate_single_asset for DRY

- Update test expectations for volume calculations

All toolkits now follow unified structure with external markdown documentation
 2025-10-30 16:22:21 +01:00
Simo
4ba44abb19 Better Prompt (#45) 
* Enhance report generation and team leader prompts with conditional output rules for empty sections.

* Update .gitignore and configuration for model adjustments; refine query classification prompt

* Enforce max response length in prompts

* Documentazione tool essenziale per il loro utilizzo da parte degli LLM.

* istruzioni chiare sui tool disponibili nei promt degli agenti

* corretto incongruenze nel report generation prompt

* convertito i promt in file markdown
 2025-10-29 15:45:05 +01:00
Copilot
c7a3199f27 Implement configurable API providers from configs.yaml (#43) 
* Implement configurable API providers from configs.yaml
* Refactor provider filtering to use WrapperHandler helper function
* Refactor API wrapper initialization to streamline configuration handling
* Refactor agent retrieval to use specific API tools directly
 2025-10-27 17:53:36 +00:00
Giacomo Bertolazzi
862525cc62 Configurazioni dell'app (#27) 
* Prompt messi in una cartella apposta
* Aggiorna importazioni demo per riflettere la nuova struttura delle cartelle API
* Aggiunto configurazione dell'applicazione
* Spostato ChatManager in app.interface
* Update README.md
* Aggiornato config per app & api
* Rinominato il modulo NewsAPI
* fix main infinite loop
* API base --> core
* pattern singleton per AppConfig.
* Estratto i tools nella loro cartella --> api/tools
* fix main KeyboardInterrupt
* update tests
* Docker & libs
* fix copilot suggestions
 2025-10-12 18:05:43 +02:00