feat: replace grpc with fibp binary protocol

[vieiralucas]

Summary

• Migrate Ruby SDK from gRPC/protobuf to the FIBP binary protocol over TCP
• Add lib/fila/fibp/ module with codec (encoding primitives), opcodes/error codes, and TCP connection manager
• Rewrite Fila::Client and Fila::Batcher to use FIBP instead of gRPC stubs
• Add admin operations: create_queue, delete_queue, get_stats, list_queues, set_config, get_config, list_config, redrive
• Add auth operations: create_api_key, revoke_api_key, list_api_keys, set_acl, get_acl
• Add new error types: AuthenticationError, ForbiddenError, NotLeaderError, QueueAlreadyExistsError, ApiKeyNotFoundError
• TLS/mTLS via Ruby OpenSSL, API key sent in FIBP handshake (not gRPC metadata)
• Leader hint redirect on NotLeaderError for consume operations
• Remove grpc and google-protobuf gem dependencies, delete all proto files
• Bump version to 0.5.0
• Net: +1214 / -854 lines

Test plan

• CI lint (rubocop) passes
• Integration tests pass against fila-server (enqueue, consume, ack, nack, batch modes)
• TLS/mTLS/API key auth tests pass
• Admin operations verified via create_queue in test helper

---

Summary by cubic

Switch the Ruby SDK from gRPC/protobuf to the FIBP binary protocol over TCP for faster, lighter I/O. Also adds transparent leader-redirect in consume, delivery/enqueue batching, and tighter error mapping.

• New Features

  • FIBP codec and TCP connection with TLS/mTLS; API key sent in handshake.
  • Background enqueue batching: :auto (default), :linger, :disabled; adds enqueue_many; close drains pending work.
  • Delivery batching in consume; automatic leader-hint reconnect on NotLeaderError.
  • Admin APIs: create/delete queue, stats, list queues, runtime config, redrive.
  • Auth APIs: create/revoke/list API keys, set/get ACL.
  • New errors: AuthenticationError, ForbiddenError, NotLeaderError, QueueAlreadyExistsError, ApiKeyNotFoundError.
  • Protocol alignment: corrected opcode constants; split stats decoding for clarity.
  • More robust single-message ack/nack result handling.

• Migration

  • Remove grpc and google-protobuf from your Gemfile; delete proto files and generation steps.
  • Initialize via Fila::Client.new(addr, ca_cert:, client_cert:, client_key:, api_key:, batch_mode:); API key is sent in the FIBP handshake.
  • Default batching is batch_mode: :auto; for strict per-call behavior (e.g., tests), use :disabled or enqueue_many.
  • Update error handling to the new error classes and messages.

^{Written for commit 2da74e7. Summary will update on new commits.}

[cubic-dev-ai]

10 issues found across 24 files

Prompt for AI agents (unresolved issues)

Check if these issues are valid — if so, understand the root cause of each and fix them. If appropriate, use sub-agents to investigate and fix each issue separately.


<file name="test/test_tls_auth.rb">

<violation number="1" location="test/test_tls_auth.rb:98">
P3: Close the temporary client in this rejection test; it currently leaks a connection when the expected exception is raised.</violation>

<violation number="2" location="test/test_tls_auth.rb:245">
P3: Close this temporary TLS client in an `ensure` block to avoid leaking connections during test execution.</violation>
</file>

<file name="test/test_helper.rb">

<violation number="1" location="test/test_helper.rb:116">
P2: Ensure the client is closed even when `create_queue` raises; otherwise failed admin calls can leak open connections.</violation>

<violation number="2" location="test/test_helper.rb:122">
P1: Close the probe client in an `ensure` block so retry failures do not leak sockets during server startup.</violation>
</file>

<file name="lib/fila/fibp/connection.rb">

<violation number="1" location="lib/fila/fibp/connection.rb:139">
P0: **TLS peer verification disabled when custom CA cert is provided.** `ctx.verify_mode` is never set in the `@ca_cert` branch, so it defaults to `VERIFY_NONE`. The custom CA cert store is populated but never actually used for verification — any server certificate will be accepted, defeating the purpose of TLS and enabling MITM attacks.

Add `verify_mode` and `verify_hostname` after setting the cert store.</violation>
</file>

<file name="lib/fila/batcher.rb">

<violation number="1" location="lib/fila/batcher.rb:157">
P2: String concatenation with `+=` in a loop is O(n²) — each iteration copies the entire accumulated buffer. Use `<<` to append in-place.</violation>
</file>

<file name="lib/fila/client.rb">

<violation number="1" location="lib/fila/client.rb:224">
P1: `decode_stats_result` is called but never defined — `get_stats` will raise `NoMethodError` at runtime. The method implementation needs to be added to the `private` section of the `Client` class.</violation>

<violation number="2" location="lib/fila/client.rb:235">
P1: `decode_list_queues_result` is called but never defined — `list_queues` will raise `NoMethodError` at runtime. The method implementation needs to be added to the `private` section of the `Client` class.</violation>

<violation number="3" location="lib/fila/client.rb:495">
P2: After a leader redirect, the `ensure` block cancels `rid` on the **new** connection instead of the old one. Cancel the subscription and nil out `rid` before reconnecting, so the ensure block doesn't send a stale cancel to the wrong server.</violation>

<violation number="4" location="lib/fila/client.rb:539">
P2: `reconnect_to` replaces `@conn` but the `Batcher` still holds a reference to the old (closed) connection. After a leader redirect during consume, subsequent enqueue calls through the batcher will fail. Consider giving the Batcher a way to update its connection reference, or having it access the client's connection indirectly.</violation>
</file>

_{Reply with feedback, questions, or to request a fix. Tag @cubic-dev-ai to re-run a review.}

[vieiralucas]

Closing unmerged. Fila is being rewritten from scratch and now ships a single client, in Rust.

Maintaining five thin SDK ports cost more than it returned — each one lagged the protocol, and this migration is the evidence: five parallel PRs, all green, all sitting unmerged for months because keeping them in step with a moving wire format was the bottleneck rather than the work itself.

The protocol has also moved past what this PR implements. Version 2 (faiscadev/fila@282984f) changes hot-path layouts: 16-byte message IDs instead of 36-char strings, u32 admin list counts, scheduling metadata on Enqueue, credit-based flow control, plus ExtendLease and delayed-Nack opcodes. Rebasing this onto v2 would be a rewrite, not a merge.

The work here was sound and the Cubic findings on it were addressed. It is preserved on the branch if anyone wants to build a Go/Python/JS/Ruby/Java client against v2 — docs/protocol.md is written to be implementable from the document alone.