chore: use a Conversation type (#3735)

crates/goose-bench/src/bench_session.rs

@@ -1,6 +1,6 @@
 use async_trait::async_trait;
 use chrono::{DateTime, Utc};
-use goose::message::Message;
+use goose::conversation::Conversation;
 
 use serde::{Deserialize, Serialize};
 use std::path::PathBuf;
@@ -19,7 +19,7 @@ pub struct BenchAgentError {
 pub trait BenchBaseSession: Send + Sync {
     async fn headless(&mut self, message: String) -> anyhow::Result<()>;
     fn session_file(&self) -> Option<PathBuf>;
-    fn message_history(&self) -> Vec<Message>;
+    fn message_history(&self) -> Conversation;
     fn get_total_token_usage(&self) -> anyhow::Result<Option<i32>>;
 }
 // struct for managing agent-session-access. to be passed to evals for benchmarking
@@ -34,7 +34,7 @@ impl BenchAgent {
         Self { session, errors }
     }
 
-    pub(crate) async fn prompt(&mut self, p: String) -> anyhow::Result<Vec<Message>> {
+    pub(crate) async fn prompt(&mut self, p: String) -> anyhow::Result<Conversation> {
         // Clear previous errors
         {
             let mut errors = self.errors.lock().await;

crates/goose-bench/src/eval_suites/core/computercontroller/script.rs

@@ -8,7 +8,7 @@ use crate::eval_suites::{
 };
 use crate::register_evaluation;
 use async_trait::async_trait;
-use goose::message::MessageContent;
+use goose::conversation::message::MessageContent;
 use rmcp::model::Role;
 use serde_json::{self, Value};
 

crates/goose-bench/src/eval_suites/core/computercontroller/web_scrape.rs

@@ -8,7 +8,7 @@ use crate::eval_suites::{
 };
 use crate::register_evaluation;
 use async_trait::async_trait;
-use goose::message::MessageContent;
+use goose::conversation::message::MessageContent;
 use rmcp::model::Role;
 use serde_json::{self, Value};
 

crates/goose-bench/src/eval_suites/core/developer/create_file.rs

@@ -8,7 +8,7 @@ use crate::eval_suites::{
 };
 use crate::register_evaluation;
 use async_trait::async_trait;
-use goose::message::MessageContent;
+use goose::conversation::message::MessageContent;
 use rmcp::model::Role;
 use serde_json::{self, Value};
 

crates/goose-bench/src/eval_suites/core/developer/list_files.rs

@@ -6,7 +6,7 @@ use crate::eval_suites::{
 };
 use crate::register_evaluation;
 use async_trait::async_trait;
-use goose::message::MessageContent;
+use goose::conversation::message::MessageContent;
 use rmcp::model::Role;
 use serde_json::{self, Value};
 

crates/goose-bench/src/eval_suites/core/developer/simple_repo_clone_test.rs

@@ -6,7 +6,7 @@ use crate::eval_suites::{
 };
 use crate::register_evaluation;
 use async_trait::async_trait;
-use goose::message::MessageContent;
+use goose::conversation::message::MessageContent;
 use rmcp::model::Role;
 use serde_json::{self, Value};
 

crates/goose-bench/src/eval_suites/core/developer_image/image.rs

@@ -6,7 +6,7 @@ use crate::eval_suites::{
 };
 use crate::register_evaluation;
 use async_trait::async_trait;
-use goose::message::MessageContent;
+use goose::conversation::message::MessageContent;
 use rmcp::model::Role;
 use serde_json::{self, Value};
 

crates/goose-bench/src/eval_suites/core/memory/save_fact.rs

@@ -8,7 +8,7 @@ use crate::eval_suites::{
 };
 use crate::register_evaluation;
 use async_trait::async_trait;
-use goose::message::MessageContent;
+use goose::conversation::message::MessageContent;
 use rmcp::model::Role;
 use serde_json::{self, Value};
 

crates/goose-bench/src/eval_suites/metrics.rs

@@ -1,6 +1,7 @@
 use crate::bench_session::BenchAgent;
 use crate::eval_suites::EvalMetricValue;
-use goose::message::{Message, MessageContent};
+use goose::conversation::message::{Message, MessageContent};
+use goose::conversation::Conversation;
 use std::collections::HashMap;
 use std::time::Instant;
 
@@ -8,7 +9,7 @@ use std::time::Instant;
 pub async fn collect_baseline_metrics(
     agent: &mut BenchAgent,
     prompt: String,
-) -> (Vec<Message>, HashMap<String, EvalMetricValue>) {
+) -> (Conversation, HashMap<String, EvalMetricValue>) {
     // Initialize metrics map
     let mut metrics = HashMap::new();
 
@@ -23,7 +24,7 @@ pub async fn collect_baseline_metrics(
                 "prompt_error".to_string(),
                 EvalMetricValue::String(format!("Error: {}", e)),
             );
-            Vec::new()
+            Conversation::new_unvalidated(Vec::new())
         }
     };
 
@@ -35,7 +36,7 @@ pub async fn collect_baseline_metrics(
     );
 
     // Count tool calls
-    let (total_tool_calls, tool_calls_by_name) = count_tool_calls(&messages);
+    let (total_tool_calls, tool_calls_by_name) = count_tool_calls(messages.messages());
     metrics.insert(
         "total_tool_calls".to_string(),
         EvalMetricValue::Integer(total_tool_calls),

crates/goose-bench/src/eval_suites/utils.rs

@@ -1,6 +1,6 @@
 use crate::bench_work_dir::BenchmarkWorkDir;
 use anyhow::{Context, Result};
-use goose::message::Message;
+use goose::conversation::message::Message;
 use std::fs::File;
 use std::io::Write;
 use std::path::PathBuf;

crates/goose-bench/src/eval_suites/vibes/blog_summary.rs

@@ -37,7 +37,7 @@ impl Evaluation for BlogSummary {
 
         // Write response to file and get the text content
         let response_text =
-            match write_response_to_file(&response, run_loc, "blog_summary_output.txt") {
+            match write_response_to_file(response.messages(), run_loc, "blog_summary_output.txt") {
                 Ok(text) => text,
                 Err(e) => {
                     println!("Warning: Failed to write blog summary output: {}", e);
@@ -59,7 +59,7 @@ impl Evaluation for BlogSummary {
         ));
 
         // Check if the fetch tool was used
-        let used_fetch_tool = crate::eval_suites::used_tool(&response, "fetch");
+        let used_fetch_tool = crate::eval_suites::used_tool(response.messages(), "fetch");
         metrics.push((
             "used_fetch_tool".to_string(),
             EvalMetricValue::Boolean(used_fetch_tool),

crates/goose-bench/src/eval_suites/vibes/flappy_bird.rs

@@ -6,7 +6,7 @@ use crate::eval_suites::{
 };
 use crate::register_evaluation;
 use async_trait::async_trait;
-use goose::message::MessageContent;
+use goose::conversation::message::MessageContent;
 use rmcp::model::Role;
 use serde_json::{self, Value};
 use std::fs;

crates/goose-bench/src/eval_suites/vibes/goose_wiki.rs

@@ -6,7 +6,7 @@ use crate::eval_suites::{
 };
 use crate::register_evaluation;
 use async_trait::async_trait;
-use goose::message::MessageContent;
+use goose::conversation::message::MessageContent;
 use rmcp::model::Role;
 use serde_json::{self, Value};
 use std::fs;

crates/goose-bench/src/eval_suites/vibes/restaurant_research.rs

@@ -50,17 +50,20 @@ Present the information in order of significance or quality. Focus specifically
         ).await;
 
         // Write response to file and get the text content
-        let response_text =
-            match write_response_to_file(&response, run_loc, "restaurant_research_output.txt") {
-                Ok(text) => text,
-                Err(e) => {
-                    println!("Warning: Failed to write restaurant research output: {}", e);
-                    // If file write fails, still continue with the evaluation
-                    response
-                        .last()
-                        .map_or_else(String::new, |msg| msg.as_concat_text())
-                }
-            };
+        let response_text = match write_response_to_file(
+            response.messages(),
+            run_loc,
+            "restaurant_research_output.txt",
+        ) {
+            Ok(text) => text,
+            Err(e) => {
+                println!("Warning: Failed to write restaurant research output: {}", e);
+                // If file write fails, still continue with the evaluation
+                response
+                    .last()
+                    .map_or_else(String::new, |msg| msg.as_concat_text())
+            }
+        };
 
         // Convert HashMap to Vec for our metrics
         let mut metrics = metrics_hashmap_to_vec(perf_metrics);
@@ -79,7 +82,7 @@ Present the information in order of significance or quality. Focus specifically
         ));
 
         // Check if the fetch tool was used
-        let used_fetch_tool = crate::eval_suites::used_tool(&response, "fetch");
+        let used_fetch_tool = crate::eval_suites::used_tool(response.messages(), "fetch");
         metrics.push((
             "used_fetch_tool".to_string(),
             EvalMetricValue::Boolean(used_fetch_tool),

crates/goose-bench/src/eval_suites/vibes/squirrel_census.rs

@@ -6,7 +6,7 @@ use crate::eval_suites::{
 };
 use crate::register_evaluation;
 use async_trait::async_trait;
-use goose::message::MessageContent;
+use goose::conversation::message::MessageContent;
 use rmcp::model::Role;
 use serde_json::{self, Value};