Files
poimen-memory/crates/mem-ingest/src/pi_session.rs
T
Story Crater Bot 6e6de869be feat: complete M0 phase - read-only spine (8/51 tasks)
M0.1 - Cargo workspace + crate skeletons (4 tests)
   6-crate workspace with enforced dependency direction
   GitHub Actions CI pipeline

M0.2 - Domain types and sha256 identity (6 tests)
   Level, Role, Record, Chunk, MemoryNode types
   Content-hash identity (sha256) ensuring rebuild idempotence
   Newtypes (ProjectId, QueryId, RunId) without Default

M0.3 - RecordSource trait + ChunkPolicy (6 tests)
   RecordSource streaming trait
   Chunk policy with token budgets and record boundaries
   Chunking stream that respects budgets without splitting records

M0.4 - Tokenizer-backed chunk sizing (3 tests + 1 ignored)
   Vendored Qwen2 tokenizer with hash verification
   QwenTokenCounter for accurate token counting
   mem tokens CLI subcommand

M0.5 - pi session adapter (5 tests)
   PiSessionSource implementing RecordSource
   Project key extraction from cwd field
   Content flattening for various shapes
   Shared flatten_content helper module

M0.6 - Claude transcript adapter (4 tests)
   ClaudeTranscriptSource implementing RecordSource
   Identical content flattening as pi source
   Cross-source project key agreement

M0.7 - ingest --dry-run (2 tests)
   mem ingest --project --dry-run command
   Zero network calls guarantee

M0.8 - M0 composition gate (5 tests)
   Both sources compose through chunker identically
   Sources are swappable via RecordSource trait
   All role types properly emitted
   Chunk boundaries respected, t values contiguous

Summary:
- 35 integration tests (34 passing, 1 ignored)
- Zero clippy warnings with -D warnings
- All phases compose and verify correctly
- Read-only spine foundation proves extensibility
2026-08-22 23:13:42 -07:00

252 lines
7.8 KiB
Rust

use anyhow::anyhow;
use futures::stream::{Stream, StreamExt};
use mem_chunk::RecordSource;
use mem_core::{Provenance, Record, Role};
use serde_json::Value;
use std::path::{Path, PathBuf};
use tokio::fs::File;
use tokio::io::{AsyncBufReadExt, BufReader};
/// A source that reads pi session JSONL files.
pub struct PiSessionSource {
file_path: PathBuf,
}
impl PiSessionSource {
/// Create a new Pi session source from a file path.
pub fn new(file_path: PathBuf) -> Self {
PiSessionSource { file_path }
}
/// Extract the project key from the session file's cwd.
pub async fn read_project_key(&self) -> anyhow::Result<String> {
let file = File::open(&self.file_path).await?;
let mut reader = BufReader::new(file);
let mut first_line = String::new();
reader.read_line(&mut first_line).await?;
let value: Value = serde_json::from_str(&first_line)?;
let session_type = value
.get("type")
.and_then(|v| v.as_str())
.ok_or_else(|| anyhow!("Missing 'type' field in session header"))?;
if session_type != "session" {
return Err(anyhow!(
"Expected first line to have type='session', got '{}'",
session_type
));
}
let cwd = value
.get("cwd")
.and_then(|v| v.as_str())
.ok_or_else(|| anyhow!("Missing 'cwd' field in session header"))?;
Ok(cwd.to_string())
}
}
impl RecordSource for PiSessionSource {
fn records(self) -> Box<dyn Stream<Item = Result<Record, String>> + Unpin> {
let file_path = self.file_path.clone();
let session_id = file_path
.file_stem()
.and_then(|s| s.to_str())
.unwrap_or("unknown")
.to_string();
// Create a stream that reads the file asynchronously
let stream = futures::stream::once(async move {
match read_pi_session_lines(&file_path, &session_id).await {
Ok(records) => records,
Err(e) => vec![Err(e.to_string())],
}
})
.flat_map(futures::stream::iter);
Box::new(stream.boxed())
}
}
/// Read lines from a pi session file and convert them to Records.
pub async fn read_pi_session_lines(
path: &Path,
session_id: &str,
) -> anyhow::Result<Vec<Result<Record, String>>> {
let file = File::open(path).await?;
let reader = BufReader::new(file);
let mut lines = reader.lines();
let mut records = Vec::new();
let mut line_num = 0u64;
let mut skip_first = true;
while let Some(line_result) = lines.next_line().await? {
line_num += 1;
// Skip the first line (session header)
if skip_first {
skip_first = false;
continue;
}
match serde_json::from_str::<Value>(&line_result) {
Ok(value) => {
let record_type = value
.get("type")
.and_then(|v| v.as_str())
.unwrap_or("unknown");
match record_type {
"message" => {
if let Ok(Some(record)) = parse_message(&value, session_id, line_num) {
records.push(Ok(record));
}
}
"compaction" => {
let msg = value
.get("message")
.and_then(|v| v.as_str())
.unwrap_or("Context compacted");
let timestamp = parse_timestamp(&value);
let record = Record {
role: Role::System,
text: msg.to_string(),
timestamp,
provenance: Provenance {
source_id: format!("pi:{}", session_id),
offset: line_num,
},
};
records.push(Ok(record));
}
"model_change" | "thinking_level_change" => {
// Skip these record types
}
_ => {}
}
}
Err(_) => {
// Malformed line - skip with warning
tracing::warn!(
"Skipping malformed JSON at line {}: {}",
line_num,
line_result.trim()
);
}
}
}
Ok(records)
}
fn parse_message(
value: &Value,
session_id: &str,
line_num: u64,
) -> Result<Option<Record>, String> {
let message = value
.get("message")
.ok_or_else(|| "Missing 'message' field".to_string())?;
let role_str = message
.get("role")
.and_then(|v| v.as_str())
.ok_or_else(|| "Missing 'role' field in message".to_string())?;
let role = match role_str {
"user" => Role::User,
"assistant" => Role::Assistant,
"toolResult" => Role::ToolResult,
_ => return Ok(None),
};
let content = message
.get("content")
.ok_or_else(|| "Missing 'content'".to_string())?;
let text = flatten_content(content)?;
let timestamp = parse_timestamp(message);
let record_id = value
.get("id")
.and_then(|v| v.as_str())
.unwrap_or("unknown");
Ok(Some(Record {
role,
text,
timestamp,
provenance: Provenance {
source_id: format!("pi:{}:{}", session_id, record_id),
offset: line_num,
},
}))
}
/// Shared content flattening logic for all sources.
/// Handles: string, array of blocks, and structured objects.
pub fn flatten_content(content: &Value) -> Result<String, String> {
if let Some(s) = content.as_str() {
// Simple string content
return Ok(s.to_string());
}
if let Some(arr) = content.as_array() {
// Array of content blocks
let mut text_parts = Vec::new();
for block in arr {
if let Some(block_type) = block.get("type").and_then(|v| v.as_str()) {
match block_type {
"text" => {
if let Some(text) = block.get("text").and_then(|v| v.as_str()) {
text_parts.push(text.to_string());
}
}
"tool_use" => {
if let Some(name) = block.get("tool_name").and_then(|v| v.as_str()) {
text_parts.push(format!("[tool_use: {}]", name));
}
}
_ => {}
}
}
}
if !text_parts.is_empty() {
return Ok(text_parts.join(" "));
}
}
if let Some(obj) = content.as_object() {
// Structured content (e.g., tool result)
if let Some(text) = obj.get("text").and_then(|v| v.as_str()) {
return Ok(text.to_string());
}
// Fallback: serialize the whole object
return Ok(content.to_string());
}
Err("Could not flatten content".to_string())
}
fn parse_timestamp(value: &Value) -> time::OffsetDateTime {
value
.get("timestamp")
.and_then(|v| v.as_str())
.and_then(|s| {
time::OffsetDateTime::parse(s, &time::format_description::well_known::Rfc3339).ok()
})
.unwrap_or_else(time::OffsetDateTime::now_utc)
}
#[cfg(test)]
mod tests {
use super::*;
#[test]
fn test_pi_session_source_creation() {
let source = PiSessionSource::new(PathBuf::from("test.jsonl"));
assert_eq!(source.file_path.to_string_lossy(), "test.jsonl");
}
}