feat: add Llama-3 renderer for Llama-3.2-1B/3B-Instruct

[hallerite]

Summary

Hand-coded Llama3Renderer for Meta's Llama-3.x chat template, plus matching parse_llama_3 parser. Initial scope: Llama-3.2-1B-Instruct and Llama-3.2-3B-Instruct (auto-routed via MODEL_RENDERER_MAP). No version bump.

How tests work without a Meta-license HF token

MODEL_RENDERER_MAP registers the canonical meta-llama/... paths so production callers auto-route. Tests load the tokenizer via the unrestricted unsloth/Llama-3.2-{1B,3B}-Instruct mirror — the chat-template SHA matches Meta's bit-for-bit and the underlying tiktoken-BPE files are identical. CI doesn't need an HF_TOKEN with Meta license access.

Implementation notes

• No <think> / reasoning channel — Llama-3 doesn't ship one. preserve_*_thinking constructor flags raise NotImplementedError if set (matches DefaultRenderer's contract for the same case).
• <|begin_of_text|> (BOS) is emitted at the start of every render; system block is always emitted with the fixed Cutting Knowledge Date / Today Date preamble even when no system message is supplied.
• date_string is a constructor kwarg, defaulting to "26 Jul 2024" (the chat template's strftime fallback) so output stays deterministic. Override per-instance for production runs that want today's date.
• tools_in_user_message defaults to True (matches chat template). Tools + JSON signatures inject into the first user message; pass False to flip to system-block mode. Both modes parity-tested.
• Single tool call per assistant message (chat template raises otherwise). Tool calls render as a JSON blob {"name": "...", "parameters": ...} inside the assistant body. Tool responses render under role ipython regardless of source role; mirrors the chat template's content | tojson branch — including the Jinja quirk that strings are iterable, so plain-string tool content gets JSON-quoted.
• parse_llama_3 detects the JSON tool-call body shape with a strict starts-with-{ + parses-as-dict-with-name check; malformed JSON falls through to content rather than dropping silently.

Tests

47 dedicated tests in tests/test_llama_3.py:

• MODEL_RENDERER_MAP shape + factory routing
• Constructor contract (default date, preserve_*_thinking rejection, tools_in_user_message toggle)
• Byte parity vs apply_chat_template across 11 conversation shapes (system + user, user-only, multi-turn, gen prompt, whitespace trimming, custom date, tools-in-user, tools-in-system, tool call round-trip, dict tool response, multiple-tool-calls rejection)
• parse_response (plain, tool call, malformed JSON fallthrough)
• Bridge contract (extends prev verbatim, matches fresh render, rejects assistant in extension, synthesises close on truncation)

Test plan

• pytest tests/test_llama_3.py — 47 cases pass on both 1B and 3B mirrors
• Full suite (pytest tests/ --ignore=tests/test_client.py) — 947 pass, 48 skipped, 1 xfailed (no regressions)
• Pre-commit hooks (ruff check + format) clean
• Maintainer with Meta-license HF_TOKEN can verify meta-llama/Llama-3.2-1B-Instruct parity directly (the unsloth mirror has been bit-verified, but a once-off canonical run is good defense in depth)

🤖 Generated with Claude Code

---

Note

Low Risk
Additive renderer and model routing with broad test coverage; scope is limited to two Instruct checkpoints and deterministic date handling for parity.

Overview
Adds a dedicated llama-3 rendering path for Llama 3.2 1B/3B Instruct, wiring meta-llama/Llama-3.2-{1B,3B}-Instruct through MODEL_RENDERER_MAP, the renderer registry, lazy exports, and Llama3RendererConfig (date_string, tools_in_user_message).

Llama3Renderer mirrors Meta’s chat template in Python: always emits BOS and a system preamble (knowledge/today dates), supports tools in the first user message or system block, single JSON tool-call per assistant turn, tool replies as ipython with Jinja-style tojson for strings, plus sampled_mask / is_content and bridge_to_next_turn. parse_llama_3 turns completions into plain text or one {"name", "parameters"} tool call.

Tests use the unsloth mirror with an explicit config; generic HF parity tests that can’t pass date_string are skipped; Llama is folded into bridge, roundtrip, and preserve-thinking matrices.

^{Reviewed by Cursor Bugbot for commit c4d9667. Bugbot is set up for automated code reviews on this repo. Configure here.}

Note

Add Llama-3 renderer for Llama-3.2-1B/3B-Instruct models

• Adds Llama3Renderer with full chat template rendering, response parsing, and turn-bridging for Llama-3.x Instruct models.
• Routes meta-llama/Llama-3.2-1B-Instruct and meta-llama/Llama-3.2-3B-Instruct model IDs to the new llama-3 renderer in MODEL_RENDERER_MAP.
• Adds Llama3RendererConfig with date_string (default '26 Jul 2024') and tools_in_user_message (default True) options; preserve_*_thinking flags are no-ops for this renderer.
• Adds parse_llama_3 to parse completions into tool calls (single JSON body with name/parameters) or plain text content.
• Enforces a single tool call per assistant message, raising ValueError on multiples; test suite skips multi-tool-call roundtrip tests for this renderer accordingly.

^{Macroscope summarized c4d9667.}

[macroscopeapp]

Approvability

Verdict: Needs human review

This PR adds a complete new Llama-3 renderer implementation (~500+ lines of new logic) with message rendering, tool call handling, and response parsing - a new feature introducing new capability that warrants human review despite good test coverage.

No code changes detected at c4d9667. Prior analysis still applies.

^{You can customize Macroscope's approvability policy. Learn more.}

[cursor]

Cursor Bugbot has reviewed your changes and found 1 potential issue.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit 47cb60b. Configure here.}