Datasets:
roanbrasil
/
devops-gitops-corpus

Dataset card Data Studio Files
xet
 Community
1
text
stringlengths
0
59.1k
### Memory Management
To make it remember conversations:
```tsx
import { LibSQLStorage } from "@voltagent/core";
const memoryStorage = new LibSQLStorage({
// configuration
});
const agent = new Agent({
// ... other config
memory: memoryStorage,
});
```
And _voilà_! Now the agent stores previous conversations.
:::tip Memory System
Memory is the most precious part of agents. It brings consistency to the user relationship and allows personalization.
:::
<ZoomableMermaid
chart={`
%%{init: {'theme':'base', 'themeVariables': {'primaryColor': '#10b981', 'primaryTextColor': '#10b981', 'primaryBorderColor': '#10b981', 'lineColor': '#10b981', 'secondaryColor': '#ecfdf5', 'tertiaryColor': '#d1fae5', 'background': '#ffffff', 'mainBkg': '#ecfdf5', 'secondBkg': '#d1fae5', 'tertiaryBkg': '#a7f3d0'}}}%%
sequenceDiagram
participant U as User
participant A as Agent
participant M as Memory
participant DB as Database
Note over U,DB: First Conversation
U->>A: Hello I'm Alex
A->>M: Save conversation
M->>DB: INSERT conversation
A->>U: Hello Alex nice to meet you
Note over U,DB: Second Conversation 1 hour later
U->>A: What did we talk about yesterday?
A->>M: Get user history
M->>DB: SELECT conversations WHERE user Alex
DB-->>M: Previous conversations
M-->>A: Past conversations
A->>U: Hello Alex! We met yesterday welcome back
`}
/>
Without memory, each conversation must start all over again. Even if the user says "we talked about this yesterday," the agent would respond with "Who are you?" Terrible experience!
### VoltOps LLM Observability Platform: Game Changer
We built VoltOps LLM Observability Platform to solve a critical problem in agent development. With it, you can visually inspect your agents:
- Preview conversation flows
- Debug calls to tools
- Track performance metrics
- Catch mistakes with ease
It's _vital_ to understand what your agent is doing in production.
:::caution Important for Production
Operating an agent in production without VoltOps LLM Observability Platform is akin to driving blindfolded. Use it definitely for debugging and optimization.
:::
## Best Practices (From My Experience)
:::important Critical Success Factors
These are practices acquired from my experience; guidelines you absolutely must follow to be successful with agent projects:
:::
**Define your agent's personality well.**
Instead of generic statements like "be helpful," get concrete. "Be a patient, friendly assistant who gives complete explanations" is so much better. Paint the personality of the agent alive.
**Choose tools wisely.**
Steer clear of the trap of supplying tools for everyone. Include only the features that you really need. Too many tools confuse the agent, too few leave the agent wanting.
**Never leave out error handling.**
APIs can fail, network outages can happen, you can hit rate limits. The agent has to handle these situations _gracefully_. Otherwise, the user experience is terrible.
**Monitor costs at all times.**
Each tool call translates to tokens, tokens translate to money. You will be shocked when you get the bill if you release without monitoring. I have been a victim of this in the past.
**Test, test, test!**
Think about edge cases. Plan what happens when the agent encounters bizarre scenarios. Take your mind to the "What if the user does something idiotic?" place and experiment.
## The Future: Where Is This Going?
**Hint: It's Not One Agent**:
Not just one agent, but agents talking to other agents. One researches, another analyzes, a third writes reports.
**More Powerful Reasoning**: