agent: raise default SPROUT_AGENT_MAX_OUTPUT_TOKENS to 32768 (#565)

Co-authored-by: Tyler Longwell <tlongwell-block@users.noreply.github.com>

Author: tlongwell-block

crates/sprout-agent/README.md

@@ -134,7 +134,7 @@ Everything is environment variables. No flags, no config files. (We are a subpro
 | `SPROUT_AGENT_SYSTEM_PROMPT` | built-in | Inline system prompt. |
 | `SPROUT_AGENT_SYSTEM_PROMPT_FILE` | — | File path. Mutually exclusive with the above. |
 | `SPROUT_AGENT_MAX_ROUNDS` | `0` | Tool-loop iteration cap. 0 = unlimited. |
-| `SPROUT_AGENT_MAX_OUTPUT_TOKENS` | `4096` | Per LLM call. |
+| `SPROUT_AGENT_MAX_OUTPUT_TOKENS` | `32768` | Per LLM call. Headroom for large tool-call inputs (e.g. file writes via heredoc); Sonnet 4 / Opus 4 cap at 64K. |
 | `SPROUT_AGENT_LLM_TIMEOUT_SECS` | `120` | |
 | `SPROUT_AGENT_TOOL_TIMEOUT_SECS` | `660` | Per-tool call timeout in seconds |
 | `SPROUT_AGENT_MAX_PARALLEL_TOOLS` | `8` | Max concurrent tool calls per turn (1 = sequential) |

crates/sprout-agent/src/config.rs

@@ -93,7 +93,7 @@ impl Config {
             base_url,
             anthropic_api_version: env_or("ANTHROPIC_API_VERSION", "2023-06-01"),
             max_rounds: parse_env("SPROUT_AGENT_MAX_ROUNDS", 0)?,
-            max_output_tokens: parse_env("SPROUT_AGENT_MAX_OUTPUT_TOKENS", 4096)?,
+            max_output_tokens: parse_env("SPROUT_AGENT_MAX_OUTPUT_TOKENS", 32_768)?,
             llm_timeout: Duration::from_secs(parse_env("SPROUT_AGENT_LLM_TIMEOUT_SECS", 120)?),
             tool_timeout: Duration::from_secs(parse_env("SPROUT_AGENT_TOOL_TIMEOUT_SECS", 660)?),
             mcp_init_timeout: Duration::from_secs(parse_env(