You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Copy file name to clipboardExpand all lines: src/assistant-kickoff/build-initial-prompt.ts
+4-3Lines changed: 4 additions & 3 deletions
Original file line number
Diff line number
Diff line change
@@ -39,14 +39,15 @@ First steps:
39
39
7. Use the LangWatch MCP to learn about prompt management and testing
40
40
8. Start implementing the core agent functionality
41
41
9. Instrument the agent with LangWatch
42
-
10. Use Scenario tests to ensure the agent is working as expected, integrate with the agent and consider it done only when all scenarios pass
43
-
11. If available from the framework, open a dev server for the user in the background and give them the url so they can play with the agent themselves
42
+
10. Use Scenario tests to ensure the agent is working as expected, integrate with the agent and consider it done only when all scenarios pass, check scenario docs on how to implement
43
+
11. If available from the framework, tell the user how to open a dev server give them the url they will be able to access so they can play with the agent themselves, don't run it for them
44
44
45
45
46
46
Remember:
47
+
- The LLM and LangWatch API keys are already available in the .env file, you don't need to set them up
47
48
- ALWAYS use LangWatch Prompt CLI for prompts (ask the MCP how)
48
49
- ALWAYS write Scenario tests for new features (ask the MCP how)
49
-
- The LLM and LangWatch API keys are already available in the .env file
50
+
- DO NOT test it "manually", always use the Scenario tests instead, do not open the dev server for the user, let them do it themselves only at the end of the implementation with everything working
0 commit comments