Reduce pub visibility in agent::api and user modules

api/: parsing module private, SamplingParams/StreamEvent/StreamResult/
AbortOnDrop/build_response_message/collect_stream to pub(crate).
Internal types (ChatRequest, ChunkChoice, Delta, etc.) to pub(crate).
StreamResult fields to pub(crate). Parsing functions to pub(super).

user/: context, subconscious, unconscious, thalamus modules private
(only chat needs pub(crate) for mind/ access).

Co-Authored-By: Proof of Concept <poc@bcachefs.org>

src/agent/api/mod.rs

@@ -7,7 +7,7 @@
 // Set POC_DEBUG=1 for verbose per-turn logging.
 
 pub mod http;
-pub mod parsing;
+mod parsing;
 mod types;
 mod openai;
 
@@ -26,7 +26,7 @@ use tokio::sync::mpsc;
 use crate::agent::tools::{self as agent_tools, summarize_args, ActiveToolCall};
 
 /// A JoinHandle that aborts its task when dropped.
-pub struct AbortOnDrop(tokio::task::JoinHandle<()>);
+pub(crate) struct AbortOnDrop(tokio::task::JoinHandle<()>);
 
 impl Drop for AbortOnDrop {
     fn drop(&mut self) {
@@ -36,7 +36,7 @@ impl Drop for AbortOnDrop {
 
 /// Sampling parameters for model generation.
 #[derive(Clone, Copy)]
-pub struct SamplingParams {
+pub(crate) struct SamplingParams {
     pub temperature: f32,
     pub top_p: f32,
     pub top_k: u32,
@@ -54,7 +54,7 @@ fn tools_to_json_str(tools: &[agent_tools::Tool]) -> String {
 
 /// Events produced by the streaming API backends.
 /// The runner reads these and decides what to display where.
-pub enum StreamEvent {
+pub(crate) enum StreamEvent {
     /// Content token from the model's response.
     Content(String),
     /// Reasoning/thinking token (internal monologue).
@@ -447,7 +447,7 @@ impl SseReader {
 /// from models that emit tool calls as text), parse them out and
 /// promote them to structured tool_calls. This way all consumers
 /// see tool calls uniformly regardless of backend.
-pub fn build_response_message(
+pub(crate) fn build_response_message(
     content: String,
     tool_calls: Vec<ToolCall>,
 ) -> Message {
@@ -574,16 +574,16 @@ pub(crate) fn log_diagnostics(
 // ---------------------------------------------------------------------------
 
 /// Result of collecting a complete response from the stream.
-pub struct StreamResult {
-    pub content: String,
-    pub tool_calls: Vec<ToolCall>,
-    pub usage: Option<Usage>,
-    pub finish_reason: Option<String>,
-    pub error: Option<String>,
+pub(crate) struct StreamResult {
+    pub(crate) content: String,
+    pub(crate) tool_calls: Vec<ToolCall>,
+    pub(crate) usage: Option<Usage>,
+    pub(crate) finish_reason: Option<String>,
+    pub(crate) error: Option<String>,
     /// Remaining display buffer (caller should flush if not in a tool call).
-    pub display_buf: String,
+    pub(crate) display_buf: String,
     /// Whether we were mid-tool-call when the stream ended.
-    pub in_tool_call: bool,
+    pub(crate) in_tool_call: bool,
 }
 
 /// Collect stream events into a complete response. Handles:
@@ -591,7 +591,7 @@ pub struct StreamResult {
 /// - Leaked tool call detection and dispatch (Qwen XML in content)
 /// - Structured tool call delta assembly (OpenAI-style)
 /// - UI forwarding (text deltas, reasoning, tool call notifications)
-pub async fn collect_stream(
+pub(crate) async fn collect_stream(
     rx: &mut mpsc::UnboundedReceiver<StreamEvent>,
     agent: &std::sync::Arc<tokio::sync::Mutex<super::Agent>>,
     active_tools: &crate::agent::tools::SharedActiveTools,