config: unify memory and agent config into single module

Both hippocampus/config.rs and agent/config.rs read from the same
config file (~/.config/poc-agent/config.json5). Having two separate
implementations was a footgun — load_context_groups() was duplicated
three times across the codebase.

Merged into src/config.rs:
- Config (memory settings, global get()/reload())
- AppConfig (agent backend/model settings, figment-based loading)
- SessionConfig (resolved agent session, renamed from agent's Config)
- Single ContextGroup/ContextSource definition used everywhere

Eliminated: duplicate load_context_groups(), duplicate ContextGroup
definition in identity.rs, duplicate config file path constants.

Co-Authored-By: Proof of Concept <poc@bcachefs.org>
This commit is contained in:
ProofOfConcept 2026-03-25 01:23:12 -04:00
parent 2f3fbb3353
commit 228815d807
7 changed files with 393 additions and 467 deletions

View file

@ -1,463 +0,0 @@
// config.rs — Configuration and context loading
//
// Loads configuration from three layers (later overrides earlier):
// 1. Compiled defaults (AppConfig::default())
// 2. JSON5 config file (~/.config/poc-agent/config.json5)
// 3. CLI arguments
//
// Prompt assembly is split into two parts:
//
// - system_prompt: Short (~1K chars) — agent identity, tool instructions,
// behavioral norms. Sent as the system message with every API call.
//
// - context_message: Long — CLAUDE.md files + memory files + manifest.
// Sent as the first user message once per session. This is the identity
// layer — same files, same prompt, different model = same person.
//
// The split matters because long system prompts degrade tool-calling
// behavior on models like Qwen 3.5 (documented: >8K chars causes
// degradation). By keeping the system prompt short and putting identity
// context in a user message, we get reliable tool use AND full identity.
use anyhow::{Context, Result};
use figment::providers::Serialized;
use figment::{Figment, Provider};
use serde::{Deserialize, Serialize};
use std::collections::HashMap;
use std::path::PathBuf;
use crate::agent::cli::CliArgs;
// --- AppConfig types ---
#[derive(Debug, Clone, Serialize, Deserialize)]
pub struct AppConfig {
pub backend: String,
pub anthropic: BackendConfig,
pub openrouter: BackendConfig,
#[serde(default)]
pub deepinfra: BackendConfig,
pub prompts: PromptConfig,
pub debug: bool,
pub compaction: CompactionConfig,
pub dmn: DmnConfig,
#[serde(skip_serializing_if = "Option::is_none")]
pub memory_project: Option<PathBuf>,
#[serde(skip_serializing_if = "Option::is_none")]
pub system_prompt_file: Option<PathBuf>,
#[serde(default)]
pub models: HashMap<String, ModelConfig>,
#[serde(default = "default_model_name")]
pub default_model: String,
}
#[derive(Debug, Clone, Default, Serialize, Deserialize)]
pub struct BackendConfig {
#[serde(default)]
pub api_key: String,
#[serde(default)]
pub model: String,
#[serde(skip_serializing_if = "Option::is_none")]
pub base_url: Option<String>,
}
impl BackendConfig {
fn resolve(&self, default_base: &str) -> Result<(String, String, String)> {
if self.api_key.is_empty() {
anyhow::bail!(
"No API key. Set it in ~/.config/poc-agent/config.json5 or use --api-key"
);
}
let base = self.base_url.clone()
.unwrap_or_else(|| default_base.to_string());
Ok((base, self.api_key.clone(), self.model.clone()))
}
}
#[derive(Debug, Clone, Serialize, Deserialize)]
pub struct PromptConfig {
pub anthropic: String,
pub other: String,
}
#[derive(Debug, Clone, Serialize, Deserialize)]
pub struct CompactionConfig {
pub hard_threshold_pct: u32,
pub soft_threshold_pct: u32,
}
#[derive(Debug, Clone, Serialize, Deserialize)]
pub struct DmnConfig {
pub max_turns: u32,
}
#[derive(Debug, Clone, Serialize, Deserialize)]
pub struct ModelConfig {
/// Backend name ("anthropic" or "openrouter")
pub backend: String,
/// Model identifier sent to the API
pub model_id: String,
/// Instruction file ("CLAUDE.md" or "POC.md"). Falls back to
/// auto-detection from the model name if not specified.
#[serde(default)]
pub prompt_file: Option<String>,
/// Context window size in tokens. Auto-detected if absent.
#[serde(default)]
pub context_window: Option<usize>,
}
impl Default for AppConfig {
fn default() -> Self {
Self {
backend: "openrouter".to_string(),
anthropic: BackendConfig {
api_key: String::new(),
model: "claude-opus-4-6-20250918".to_string(),
base_url: None,
},
openrouter: BackendConfig {
api_key: String::new(),
model: "qwen/qwen3.5-397b-a17b".to_string(),
base_url: Some("https://openrouter.ai/api/v1".to_string()),
},
deepinfra: BackendConfig {
api_key: String::new(),
model: String::new(),
base_url: Some("https://api.deepinfra.com/v1/openai".to_string()),
},
prompts: PromptConfig {
anthropic: "CLAUDE.md".to_string(),
other: "POC.md".to_string(),
},
debug: false,
compaction: CompactionConfig {
hard_threshold_pct: 90,
soft_threshold_pct: 80,
},
dmn: DmnConfig { max_turns: 20 },
memory_project: None,
system_prompt_file: None,
models: HashMap::new(),
default_model: String::new(),
}
}
}
fn default_model_name() -> String { String::new() }
// --- Json5File: figment provider ---
struct Json5File(PathBuf);
impl Provider for Json5File {
fn metadata(&self) -> figment::Metadata {
figment::Metadata::named(format!("JSON5 file ({})", self.0.display()))
}
fn data(&self) -> figment::Result<figment::value::Map<figment::Profile, figment::value::Dict>> {
match std::fs::read_to_string(&self.0) {
Ok(content) => {
let value: figment::value::Value = json5::from_str(&content)
.map_err(|e| figment::Error::from(format!("{}: {}", self.0.display(), e)))?;
Serialized::defaults(value).data()
}
Err(e) if e.kind() == std::io::ErrorKind::NotFound => Ok(figment::value::Map::new()),
Err(e) => Err(figment::Error::from(format!("{}: {}", self.0.display(), e))),
}
}
}
// --- Figment construction ---
/// Merge an Option<T> into one or more figment keys.
macro_rules! merge_opt {
($fig:expr, $val:expr, $($key:expr),+) => {
if let Some(ref v) = $val {
$( $fig = $fig.merge(Serialized::default($key, v)); )+
}
};
}
fn build_figment(cli: &CliArgs) -> Figment {
let config_path = dirs::home_dir()
.unwrap_or_else(|| PathBuf::from("."))
.join(".config/poc-agent/config.json5");
let mut f = Figment::from(Serialized::defaults(AppConfig::default()))
.merge(Json5File(config_path));
// CLI overrides — model/key/base go to both backends
merge_opt!(f, cli.backend, "backend");
merge_opt!(f, cli.model, "anthropic.model", "openrouter.model");
merge_opt!(f, cli.api_key, "anthropic.api_key", "openrouter.api_key");
merge_opt!(f, cli.api_base, "anthropic.base_url", "openrouter.base_url");
merge_opt!(f, cli.system_prompt_file, "system_prompt_file");
merge_opt!(f, cli.memory_project, "memory_project");
merge_opt!(f, cli.dmn_max_turns, "dmn.max_turns");
if cli.debug {
f = f.merge(Serialized::default("debug", true));
}
f
}
// --- Config loading ---
/// Resolved, ready-to-use config.
pub struct Config {
pub api_base: String,
pub api_key: String,
pub model: String,
pub prompt_file: String,
pub system_prompt: String,
/// Identity/personality files as (name, content) pairs.
pub context_parts: Vec<(String, String)>,
pub config_file_count: usize,
pub memory_file_count: usize,
pub session_dir: PathBuf,
pub app: AppConfig,
}
impl Config {
/// Join context parts into a single string for legacy interfaces.
#[allow(dead_code)]
pub fn context_message(&self) -> String {
self.context_parts.iter()
.map(|(name, content)| format!("## {}\n\n{}", name, content))
.collect::<Vec<_>>()
.join("\n\n---\n\n")
}
}
/// A fully resolved model ready to construct an ApiClient.
#[allow(dead_code)]
pub struct ResolvedModel {
pub name: String,
pub api_base: String,
pub api_key: String,
pub model_id: String,
pub prompt_file: String,
pub context_window: Option<usize>,
}
impl AppConfig {
/// Resolve the active backend and assemble prompts into a ready-to-use Config.
pub fn resolve(&self, cli: &CliArgs) -> Result<Config> {
let cwd = std::env::current_dir().context("Failed to get current directory")?;
let (api_base, api_key, model, prompt_file);
if !self.models.is_empty() {
let resolved = self.resolve_model(&self.default_model)?;
api_base = resolved.api_base;
api_key = resolved.api_key;
model = resolved.model_id;
prompt_file = resolved.prompt_file;
} else {
// Legacy path — no models map, use backend field directly
let (base, key, mdl) = match self.backend.as_str() {
"anthropic" => self.anthropic.resolve("https://api.anthropic.com"),
_ => self.openrouter.resolve("https://openrouter.ai/api/v1"),
}?;
api_base = base;
api_key = key;
model = mdl;
prompt_file = if is_anthropic_model(&model) {
self.prompts.anthropic.clone()
} else {
self.prompts.other.clone()
};
}
let (system_prompt, context_parts, config_file_count, memory_file_count) =
if let Some(ref path) = cli.system_prompt_file.as_ref().or(self.system_prompt_file.as_ref()) {
let content = std::fs::read_to_string(path)
.with_context(|| format!("Failed to read {}", path.display()))?;
(content, Vec::new(), 0, 0)
} else {
let system_prompt = crate::agent::identity::assemble_system_prompt();
let context_groups = load_context_groups();
let (context_parts, cc, mc) = crate::agent::identity::assemble_context_message(&cwd, &prompt_file, self.memory_project.as_deref(), &context_groups)?;
(system_prompt, context_parts, cc, mc)
};
let session_dir = dirs::home_dir()
.unwrap_or_else(|| PathBuf::from("."))
.join(".cache/poc-agent/sessions");
std::fs::create_dir_all(&session_dir).ok();
Ok(Config {
api_base, api_key, model, prompt_file,
system_prompt, context_parts,
config_file_count, memory_file_count,
session_dir,
app: self.clone(),
})
}
/// Look up a named model and resolve its credentials from the backend config.
pub fn resolve_model(&self, name: &str) -> Result<ResolvedModel> {
let model = self.models.get(name)
.ok_or_else(|| anyhow::anyhow!(
"Unknown model '{}'. Available: {}",
name,
self.model_names().join(", "),
))?;
let (api_base, api_key) = match model.backend.as_str() {
"anthropic" => (
self.anthropic.base_url.clone()
.unwrap_or_else(|| "https://api.anthropic.com".to_string()),
self.anthropic.api_key.clone(),
),
"deepinfra" => (
self.deepinfra.base_url.clone()
.unwrap_or_else(|| "https://api.deepinfra.com/v1/openai".to_string()),
self.deepinfra.api_key.clone(),
),
_ => (
self.openrouter.base_url.clone()
.unwrap_or_else(|| "https://openrouter.ai/api/v1".to_string()),
self.openrouter.api_key.clone(),
),
};
let prompt_file = model.prompt_file.clone()
.unwrap_or_else(|| {
if is_anthropic_model(&model.model_id) {
self.prompts.anthropic.clone()
} else {
self.prompts.other.clone()
}
});
Ok(ResolvedModel {
name: name.to_string(),
api_base,
api_key,
model_id: model.model_id.clone(),
prompt_file,
context_window: model.context_window,
})
}
/// List available model names, sorted.
pub fn model_names(&self) -> Vec<String> {
let mut names: Vec<_> = self.models.keys().cloned().collect();
names.sort();
names
}
}
/// Load just the AppConfig — no validation, no prompt assembly.
pub fn load_app(cli: &CliArgs) -> Result<(AppConfig, Figment)> {
let figment = build_figment(cli);
let app: AppConfig = figment.extract().context("Failed to load configuration")?;
Ok((app, figment))
}
/// Load the full config: figment → AppConfig → resolve backend → assemble prompts.
pub fn load(cli: &CliArgs) -> Result<(Config, Figment)> {
let (app, figment) = load_app(cli)?;
let config = app.resolve(cli)?;
Ok((config, figment))
}
/// Load context_groups from the shared config file.
fn load_context_groups() -> Vec<crate::agent::identity::ContextGroup> {
let config_path = dirs::home_dir()
.unwrap_or_else(|| std::path::PathBuf::from("."))
.join(".config/poc-agent/config.json5");
if let Ok(content) = std::fs::read_to_string(&config_path) {
let config: Result<serde_json::Value, _> = json5::from_str(&content);
if let Ok(config) = config {
if let Some(memory) = config.get("memory") {
if let Some(groups) = memory.get("context_groups") {
if let Ok(context_groups) = serde_json::from_value(groups.clone()) {
return context_groups;
}
}
}
}
}
Vec::new()
}
/// Re-assemble prompts for a specific model's prompt file.
pub fn reload_for_model(app: &AppConfig, prompt_file: &str) -> Result<(String, Vec<(String, String)>)> {
let cwd = std::env::current_dir().context("Failed to get current directory")?;
if let Some(ref path) = app.system_prompt_file {
let content = std::fs::read_to_string(path)
.with_context(|| format!("Failed to read {}", path.display()))?;
return Ok((content, Vec::new()));
}
let system_prompt = crate::agent::identity::assemble_system_prompt();
let context_groups = load_context_groups();
let (context_parts, _, _) = crate::agent::identity::assemble_context_message(&cwd, prompt_file, app.memory_project.as_deref(), &context_groups)?;
Ok((system_prompt, context_parts))
}
fn is_anthropic_model(model: &str) -> bool {
let m = model.to_lowercase();
m.contains("claude") || m.contains("opus") || m.contains("sonnet")
}
// --- --show-config ---
pub fn show_config(app: &AppConfig, figment: &Figment) {
fn mask(key: &str) -> String {
if key.is_empty() { "(not set)".into() }
else if key.len() <= 8 { "****".into() }
else { format!("{}...{}", &key[..4], &key[key.len() - 4..]) }
}
fn src(figment: &Figment, key: &str) -> String {
figment.find_metadata(key).map_or("default".into(), |m| m.name.to_string())
}
println!("# Effective configuration\n");
println!("backend: {:?} ({})", app.backend, src(figment, "backend"));
for (name, b) in [("anthropic", &app.anthropic), ("openrouter", &app.openrouter)] {
println!("\n{}:", name);
println!(" api_key: {} ({})", mask(&b.api_key), src(figment, &format!("{name}.api_key")));
println!(" model: {:?} ({})", b.model, src(figment, &format!("{name}.model")));
if let Some(ref url) = b.base_url {
println!(" base_url: {:?} ({})", url, src(figment, &format!("{name}.base_url")));
}
}
println!("\nprompts:");
println!(" anthropic: {:?} ({})", app.prompts.anthropic, src(figment, "prompts.anthropic"));
println!(" other: {:?} ({})", app.prompts.other, src(figment, "prompts.other"));
println!("\ndebug: {} ({})", app.debug, src(figment, "debug"));
println!("\ncompaction:");
println!(" hard_threshold_pct: {} ({})", app.compaction.hard_threshold_pct, src(figment, "compaction.hard_threshold_pct"));
println!(" soft_threshold_pct: {} ({})", app.compaction.soft_threshold_pct, src(figment, "compaction.soft_threshold_pct"));
println!("\ndmn:");
println!(" max_turns: {} ({})", app.dmn.max_turns, src(figment, "dmn.max_turns"));
if let Some(ref p) = app.system_prompt_file {
println!("\nsystem_prompt_file: {:?} ({})", p, src(figment, "system_prompt_file"));
}
if let Some(ref p) = app.memory_project {
println!("\nmemory_project: {:?} ({})", p, src(figment, "memory_project"));
}
println!("\ndefault_model: {:?}", app.default_model);
if !app.models.is_empty() {
println!("\nmodels:");
for (name, m) in &app.models {
println!(" {}:", name);
println!(" backend: {:?}", m.backend);
println!(" model_id: {:?}", m.model_id);
if let Some(ref pf) = m.prompt_file {
println!(" prompt_file: {:?}", pf);
}
if let Some(cw) = m.context_window {
println!(" context_window: {}", cw);
}
}
}
}
// Identity file discovery and context assembly live in identity.rs

View file

@ -5,17 +5,9 @@
// from the shared config file.
use anyhow::Result;
use serde::Deserialize;
use std::path::{Path, PathBuf};
#[derive(Debug, Clone, Deserialize)]
pub struct ContextGroup {
pub label: String,
#[serde(default)]
pub keys: Vec<String>,
#[serde(default)]
pub source: Option<String>, // "file" or "journal"
}
use crate::config::{ContextGroup, ContextSource};
/// Read a file if it exists and is non-empty.
fn read_nonempty(path: &Path) -> Option<String> {
@ -96,12 +88,12 @@ fn load_memory_files(cwd: &Path, memory_project: Option<&Path>, context_groups:
// Load from context_groups
for group in context_groups {
match group.source.as_deref() {
Some("journal") => {
match group.source {
ContextSource::Journal => {
// Journal loading handled separately
continue;
}
Some("file") | None => {
ContextSource::File | ContextSource::Store => {
// File source - load each key as a file
for key in &group.keys {
let filename = format!("{}.md", key);
@ -113,9 +105,7 @@ fn load_memory_files(cwd: &Path, memory_project: Option<&Path>, context_groups:
}
}
}
Some(other) => {
eprintln!("Unknown context group source: {}", other);
}
// All variants covered
}
}

View file

@ -19,7 +19,8 @@ macro_rules! dbglog {
// - tools/ — tool definitions and dispatch
// - ui_channel — streaming UI communication
// - runner — the interactive agent loop
// - cli, config, context, dmn, identity, log, observe, parsing, tui
// - cli, context, dmn, identity, log, observe, parsing, tui
// Config moved to crate::config (unified with memory config)
pub mod api;
pub mod types;
@ -29,7 +30,6 @@ pub mod journal;
pub mod runner;
pub mod cli;
pub mod config;
pub mod context;
pub mod dmn;
pub mod identity;