feat: complete M0.1-M0.4 phases

M0.1 - Cargo workspace + crate skeletons
  - 6-crate workspace with correct dependency direction
  - CI/CD pipeline with GitHub Actions
  - Integration tests verifying build and dependency structure

M0.2 - Domain types and sha256 identity
  - Level (L0, L1, L2) enum with proper serde formatting
  - Role enum (User, Assistant, ToolResult, System)
  - Record, Chunk, and MemoryNode domain types
  - Content-hash identity system ensuring rebuild idempotence
  - Newtypes (ProjectId, QueryId, RunId) with validation
  - Round-trip serde tests for all types

M0.3 - RecordSource trait + ChunkPolicy
  - RecordSource trait for streaming record sources
  - Chunk policy with token budgets and boundary modes
  - TokenCounter trait with CharsOverFourCounter stub
  - Chunking stream that respects budgets without splitting records
  - VecSource for testing
  - Integration tests verifying lossless chunking and budget adherence

M0.4 - Tokenizer-backed chunk sizing
  - Vendored Qwen2 tokenizer with hash verification
  - QwenTokenCounter implementing proper token counting
  - Hash guard that fails on modified tokenizer
  - mem tokens CLI subcommand for token counting
  - Integration tests with known string counts, hash guards, and budget verification

Total: 19 integration tests passing, all phases verified to compose correctly
Workspace builds cleanly with no clippy warnings
This commit is contained in:
Story Crater Bot
2026-08-22 23:13:42 -07:00
parent a163c03619
commit 33b7150f56
27 changed files with 2216 additions and 1 deletions
+16
View File
@@ -0,0 +1,16 @@
[package]
name = "mem-core"
version = "0.1.0"
edition = "2021"
[dependencies]
tokio = { workspace = true }
futures = { workspace = true }
serde = { workspace = true }
serde_json = { workspace = true }
anyhow = { workspace = true }
thiserror = { workspace = true }
sha2 = { workspace = true }
tracing = { workspace = true }
hex = "0.4"
time = { version = "0.3", features = ["serde", "formatting", "parsing", "macros"] }
+402
View File
@@ -0,0 +1,402 @@
use serde::{Deserialize, Serialize};
use sha2::{Digest, Sha256};
use std::fmt;
use std::str::FromStr;
use time::OffsetDateTime;
/// Memory node level in the hierarchy.
/// Closed. L0 evidence, L1 per-query memory, L2 project synthesis.
#[derive(Clone, Copy, Debug, PartialEq, Eq, Hash, Serialize, Deserialize)]
#[serde(rename_all = "UPPERCASE")]
pub enum Level {
#[serde(rename = "L0")]
L0,
#[serde(rename = "L1")]
L1,
#[serde(rename = "L2")]
L2,
}
impl fmt::Display for Level {
fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
match self {
Level::L0 => write!(f, "L0"),
Level::L1 => write!(f, "L1"),
Level::L2 => write!(f, "L2"),
}
}
}
impl FromStr for Level {
type Err = String;
fn from_str(s: &str) -> Result<Self, Self::Err> {
match s {
"L0" => Ok(Level::L0),
"L1" => Ok(Level::L1),
"L2" => Ok(Level::L2),
_ => Err(format!("Invalid level: {}", s)),
}
}
}
/// The role of a record in a conversation.
#[derive(Clone, Copy, Debug, PartialEq, Eq, Hash, Serialize, Deserialize)]
#[serde(rename_all = "PascalCase")]
pub enum Role {
User,
Assistant,
ToolResult,
System,
}
/// Source identification and offset.
#[derive(Clone, Debug, PartialEq, Eq, Hash, Serialize, Deserialize)]
pub struct Provenance {
pub source_id: String,
pub offset: u64,
}
/// A normalized unit from any source.
/// Adapters produce these; nothing downstream learns whether it came from pi,
/// claude, or a socket.
#[derive(Clone, Debug, Serialize, Deserialize)]
pub struct Record {
pub role: Role,
pub text: String,
#[serde(with = "time::serde::rfc3339")]
pub timestamp: OffsetDateTime,
pub provenance: Provenance,
}
/// Content hash as a hex string.
#[derive(Clone, Copy, Debug, PartialEq, Eq, Hash, Serialize, Deserialize, PartialOrd, Ord)]
pub struct Sha256Hash([u8; 32]);
impl Sha256Hash {
/// Create a hash from a byte array.
pub fn from_bytes(bytes: [u8; 32]) -> Self {
Sha256Hash(bytes)
}
/// Create a hash from a hex string.
pub fn from_hex(hex: &str) -> Result<Self, String> {
if hex.len() != 64 {
return Err("Hash must be 64 hex characters".to_string());
}
let mut bytes = [0u8; 32];
for (i, chunk) in hex.as_bytes().chunks(2).enumerate() {
bytes[i] = u8::from_str_radix(std::str::from_utf8(chunk).unwrap(), 16)
.map_err(|_| "Invalid hex character".to_string())?;
}
Ok(Sha256Hash(bytes))
}
/// Convert to hex string.
pub fn to_hex(&self) -> String {
hex::encode(self.0)
}
/// Get the raw bytes.
pub fn as_bytes(&self) -> &[u8] {
&self.0
}
}
impl fmt::Display for Sha256Hash {
fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
write!(f, "{}", self.to_hex())
}
}
/// Newtype wrappers with no Default.
#[derive(Clone, Debug, PartialEq, Eq, Hash, Serialize, Deserialize)]
pub struct ProjectId(String);
impl ProjectId {
pub fn new(id: String) -> Result<Self, String> {
if id.is_empty() {
return Err("ProjectId cannot be empty".to_string());
}
Ok(ProjectId(id))
}
pub fn as_str(&self) -> &str {
&self.0
}
}
#[derive(Clone, Debug, PartialEq, Eq, Hash, Serialize, Deserialize)]
pub struct QueryId(String);
impl QueryId {
pub fn new(id: String) -> Result<Self, String> {
if id.is_empty() {
return Err("QueryId cannot be empty".to_string());
}
Ok(QueryId(id))
}
pub fn as_str(&self) -> &str {
&self.0
}
}
#[derive(Clone, Debug, PartialEq, Eq, Hash, Serialize, Deserialize)]
pub struct RunId(String);
impl RunId {
pub fn new(id: String) -> Result<Self, String> {
if id.is_empty() {
return Err("RunId cannot be empty".to_string());
}
Ok(RunId(id))
}
pub fn as_str(&self) -> &str {
&self.0
}
}
/// One or more Records, under the token budget, never split mid-Record.
#[derive(Clone, Debug, Serialize, Deserialize)]
pub struct Chunk {
pub t: u32, // 1-based turn index within a run
pub records: Vec<Record>,
pub tokens: usize,
#[serde(skip)]
sha256: Option<Sha256Hash>,
}
impl Chunk {
pub fn new(t: u32, records: Vec<Record>, tokens: usize) -> Self {
Chunk {
t,
records,
tokens,
sha256: None,
}
}
/// Compute canonical hash over concatenated record texts and their provenance.
/// Must not include timestamp or run_id to ensure rebuild idempotence.
pub fn content_hash(&mut self) -> Sha256Hash {
if let Some(hash) = self.sha256 {
return hash;
}
let mut hasher = Sha256::new();
// Concatenate record texts and provenance
for record in &self.records {
hasher.update(record.role.to_string().as_bytes());
hasher.update(b"\x00");
hasher.update(record.text.as_bytes());
hasher.update(b"\x00");
hasher.update(record.provenance.source_id.as_bytes());
hasher.update(b"\x00");
hasher.update(record.provenance.offset.to_le_bytes());
hasher.update(b"\x00");
}
let bytes: [u8; 32] = hasher.finalize().into();
let hash = Sha256Hash::from_bytes(bytes);
self.sha256 = Some(hash);
hash
}
pub fn sha256(&mut self) -> Sha256Hash {
self.content_hash()
}
}
impl fmt::Display for Role {
fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
match self {
Role::User => write!(f, "User"),
Role::Assistant => write!(f, "Assistant"),
Role::ToolResult => write!(f, "ToolResult"),
Role::System => write!(f, "System"),
}
}
}
/// A memory node at any level.
#[derive(Clone, Debug, Serialize, Deserialize)]
pub struct MemoryNode {
pub level: Level,
pub project: ProjectId,
pub query_id: Option<QueryId>, // None at L2
pub run_id: RunId,
pub t: u32,
pub text: String,
#[serde(skip)]
sha256: Option<Sha256Hash>,
pub parents: Vec<Sha256Hash>,
}
impl MemoryNode {
pub fn new(
level: Level,
project: ProjectId,
query_id: Option<QueryId>,
run_id: RunId,
t: u32,
text: String,
parents: Vec<Sha256Hash>,
) -> Self {
MemoryNode {
level,
project,
query_id,
run_id,
t,
text,
sha256: None,
parents,
}
}
/// Compute canonical hash over (level, project, query_id, text).
/// Must not include timestamp or run_id to ensure rebuild idempotence.
pub fn content_hash(&mut self) -> Sha256Hash {
if let Some(hash) = self.sha256 {
return hash;
}
let mut hasher = Sha256::new();
hasher.update(self.level.to_string().as_bytes());
hasher.update(b"\x00");
hasher.update(self.project.as_str().as_bytes());
hasher.update(b"\x00");
if let Some(query_id) = &self.query_id {
hasher.update(query_id.as_str().as_bytes());
}
hasher.update(b"\x00");
hasher.update(self.text.as_bytes());
let bytes: [u8; 32] = hasher.finalize().into();
let hash = Sha256Hash::from_bytes(bytes);
self.sha256 = Some(hash);
hash
}
pub fn sha256(&mut self) -> Sha256Hash {
self.content_hash()
}
}
#[cfg(test)]
mod tests {
use super::*;
use time::macros::datetime;
#[test]
fn test_level_serialization() {
assert_eq!(serde_json::to_string(&Level::L0).unwrap(), "\"L0\"");
assert_eq!(serde_json::to_string(&Level::L1).unwrap(), "\"L1\"");
assert_eq!(serde_json::to_string(&Level::L2).unwrap(), "\"L2\"");
}
#[test]
fn test_level_round_trip() {
for level in &[Level::L0, Level::L1, Level::L2] {
let json = serde_json::to_string(level).unwrap();
let deserialized: Level = serde_json::from_str(&json).unwrap();
assert_eq!(level, &deserialized);
}
}
#[test]
fn test_role_round_trip() {
for role in &[Role::User, Role::Assistant, Role::ToolResult, Role::System] {
let json = serde_json::to_string(role).unwrap();
let deserialized: Role = serde_json::from_str(&json).unwrap();
assert_eq!(role, &deserialized);
}
}
#[test]
fn test_sha256_hash_round_trip() {
let original = Sha256Hash::from_bytes([1; 32]);
let json = serde_json::to_string(&original).unwrap();
let deserialized: Sha256Hash = serde_json::from_str(&json).unwrap();
assert_eq!(original, deserialized);
}
#[test]
fn test_project_id_creation() {
let id = ProjectId::new("project1".to_string()).unwrap();
assert_eq!(id.as_str(), "project1");
let result = ProjectId::new("".to_string());
assert!(result.is_err());
}
#[test]
fn test_record_round_trip() {
let record = Record {
role: Role::User,
text: "Hello, world!".to_string(),
timestamp: datetime!(2024-08-20 12:00:00 UTC),
provenance: Provenance {
source_id: "session1".to_string(),
offset: 0,
},
};
let json = serde_json::to_string(&record).unwrap();
let deserialized: Record = serde_json::from_str(&json).unwrap();
assert_eq!(record.role, deserialized.role);
assert_eq!(record.text, deserialized.text);
assert_eq!(record.provenance, deserialized.provenance);
}
#[test]
fn test_chunk_round_trip() {
let record = Record {
role: Role::User,
text: "Hello".to_string(),
timestamp: datetime!(2024-08-20 12:00:00 UTC),
provenance: Provenance {
source_id: "session1".to_string(),
offset: 0,
},
};
let chunk = Chunk {
t: 1,
records: vec![record],
tokens: 2,
sha256: None,
};
let json = serde_json::to_string(&chunk).unwrap();
let deserialized: Chunk = serde_json::from_str(&json).unwrap();
assert_eq!(chunk.t, deserialized.t);
assert_eq!(chunk.tokens, deserialized.tokens);
}
#[test]
fn test_memory_node_round_trip() {
let node = MemoryNode {
level: Level::L0,
project: ProjectId::new("p1".to_string()).unwrap(),
query_id: Some(QueryId::new("q1".to_string()).unwrap()),
run_id: RunId::new("r1".to_string()).unwrap(),
t: 1,
text: "test".to_string(),
sha256: None,
parents: vec![],
};
let json = serde_json::to_string(&node).unwrap();
let deserialized: MemoryNode = serde_json::from_str(&json).unwrap();
assert_eq!(node.level, deserialized.level);
assert_eq!(node.t, deserialized.t);
assert_eq!(node.text, deserialized.text);
}
}
+5
View File
@@ -0,0 +1,5 @@
pub mod domain;
pub use domain::{
Chunk, Level, MemoryNode, Provenance, Record, Role, ProjectId, QueryId, RunId, Sha256Hash,
};