Improve agentapi module abstraction for configuration and flags #696

Summary

Currently, starting agentapi and passing flags is the responsibility of individual agent modules rather than a shared base agentapi module. This leads to repeated code across modules.

Current State

Each module implements its own agentapi invocation with various flags:

agentapi server --type opencode --term-width 67 --term-height 1190 -- opencode "${ARGS[@]}"
agentapi server --type amp --term-width=67 --term-height=1190 -- bash -c "echo \"$PROMPT\" | amp" "${ARGS[@]}"
agentapi server --type claude --term-width 67 --term-height 1190 -- claude "${ARGS[@]}"
agentapi server --term-width=67 --term-height=1190 -- aider --model $ARG_MODEL --yes-always

Proposed Improvement

1. A configuration agentapi module could provide a standardized agentapi command that operates in a certain way (e.g., with regards to state persistence)

2. Common configuration options could be passed as module variables:

   • --type
   • --term-width
   • --term-height
   • --state-file

3. Agent-specific startup logic can be written to a shell script that gets executed by the agentapi server:

   agentapi server --type claude --state-file="/tmp/agent.state.json" --term-width 67 --term-height 1190 -- ./start-agent.sh

Related

• AgentAPI state persistence feature (feat: implement state persistence agentapi#177)

Created on behalf of @work-jaykumar