feat: /conversations backing store — memory and Pebble backends

[elevran]

Scope

Implement the ConversationStore — the durable backing store for the /conversations endpoint. This is an internal package (internal/convstore) consumed by both cmd/conversationd (direct) and the proxy (via the convstore client or an embedded store in single-binary mode).

Design decisions

What a conversation is

A conversation is a named container for a sequence of messages. It is a higher-level concept than a response chain: a single conversation can span many response turns, each of which may have been stored in chainstore independently. The conversation store holds:

• Conversation metadata: id, created_at, updated_at, title (optional), metadata (key-value)
• Message list: ordered append-only log of Message values. Each message contains at minimum {role, content[]} (mirrors OpenAI Conversations API message objects).

ID allocation

Conversation IDs (conv_*) are minted by cmd/conversationd at POST /conversations time — not by the proxy. This mirrors how response IDs are assigned by the inference server. The proxy may pre-request an ID via POST /conversations before inference, or post-allocate after.

Storage key scheme (Pebble)

Key prefix	Value
conv:meta:{id}	Serialised ConversationMeta proto/JSON
conv:msg:{id}:{seq:8}	Serialised Message (seq is a zero-padded uint64 for lexicographic ordering)
conv:lru:{bucket}:{id}	Tombstone for LRU eviction index (mirrors chainstore bucket design)

Interface

type ConversationStore interface {
    Create(ctx context.Context, meta ConversationMeta) (ConversationMeta, error)
    Get(ctx context.Context, id string) (ConversationMeta, error)
    Delete(ctx context.Context, id string) error
    AppendMessages(ctx context.Context, id string, messages []Message) (nextSeq uint64, err error)
    ListMessages(ctx context.Context, id string, afterSeq uint64, limit int) ([]Message, error)
    Ping(ctx context.Context) error
    Close() error
}

Config mirrors chainstore.Config

type Config struct {
    MaxConversations int64
    MaxMessagesPerConversation int
    MaxBytes         int64
    TTL              time.Duration
    TTLInterval      time.Duration
    EvictionInterval time.Duration
    BucketDuration   time.Duration
    Clock            Clock
    Log              *slog.Logger
    Backend          Backend  // "memory" or pebble.Open()
    Registerer       prometheus.Registerer
}

Backends

• In-memory (convstore/memory): sync.RWMutex-protected maps; suitable for tests and the store: false-style use case. No TTL persistence.
• Pebble (convstore/pebble): reuses the same Pebble DB opened by chainstore (passed in via Backend) or opens its own. Separate key prefix ensures no collision with chainstore keys.

Capacity management

• LRU eviction when MaxConversations or MaxBytes is exceeded (same bucket-based eviction as chainstore)
• Background TTL reaper (configurable interval)
• MaxMessagesPerConversation cap: AppendMessages returns ErrConversationFull when exceeded

Acceptance criteria

• internal/convstore package compiles with both in-memory and Pebble backends
• ConversationStore interface covers Create / Get / Delete / AppendMessages / ListMessages
• In-memory backend passes a shared test suite (convstore/testutil)
• Pebble backend passes the same test suite
• TTL eviction test: create conversations with short TTL, advance clock, verify reaped
• Capacity eviction test: exceed MaxConversations, verify oldest evicted
• MaxMessagesPerConversation cap returns correct error
• make presubmit passes

Blocked by: #77