verifiers 0.3.2.dev37__py3-none-any.whl → 0.3.2.dev39__py3-none-any.whl

verifiers/v1/harnesses/prime_agent/harness.py

@@ -19,10 +19,10 @@ logger = logging.getLogger(__name__)
 GITHUB_RELEASE_URL = (
     "https://github.com/PrimeIntellect-ai/prime-agent/releases/download"
 )
-PRIME_AGENT_COMMIT: Literal["514633727bf26d74f39f3119c2b0e31a5ceb2a9d"] = (
-    "514633727bf26d74f39f3119c2b0e31a5ceb2a9d"
+PRIME_AGENT_COMMIT: Literal["81ae3cb34d27d38ee37f9e205a1e73694993b344"] = (
+    "81ae3cb34d27d38ee37f9e205a1e73694993b344"
 )
-PRIME_AGENT_VERSION = "0.8.1"
+PRIME_AGENT_VERSION = "0.9.1"
 PRIME_AGENT_DIR = "/var/tmp/vf-prime-agent"
 STATE_ROOT = "/tmp/vf-prime-agent-runs"
 SKILLS_DIR = ".agents/skills"
@@ -51,10 +51,10 @@ for tarball in "$agent_tarball" "$ai_tarball" "$core_tarball" "$tui_tarball"; do
         "$release_url/$tarball" -o "$download_dir/$tarball"
 done
 printf '%s  %s\n' \
-    '46c24db1782dd31adc35d5c6cbcc75564faba6ced3bf2ccf03d836ee77134475' "$agent_tarball" \
-    'f6c3bdb6093bc24a327546fe865ef9a4a172c734fcd4c4093e30c19476f0134d' "$ai_tarball" \
-    '0cc3660953545f8ac9a7e704fcb9875f954d58c3085304080ef615c280aa5748' "$core_tarball" \
-    'bd07bccee0ca495565b1d62e9411f3fdebe49e3dfa52870564f08af5e61fde15' "$tui_tarball" \
+    '573bce0cd004fc62052e9a924089941b7f39266ab71e66a94c85a1f9d35835ba' "$agent_tarball" \
+    '11b5b4cf67b6bb2d3420a44fb69181bc9d94d81e69a2b4fde07eb9c99f5faf4f' "$ai_tarball" \
+    'fb6f3a5dcc8b69c5eeb3beff722b5e0f09885c14849db50bc1d7c0f1d064151c' "$core_tarball" \
+    '4f3eaca2814944d3993073e0b88c0bde54a641ddb5132fbe107e392b997e38ec' "$tui_tarball" \
     > "$download_dir/SHA256SUMS"
 (cd "$download_dir" && sha256sum -c SHA256SUMS)
 mkdir "$download_dir/package-root"
@@ -87,7 +87,7 @@ PRIME_AGENT_BOOTSTRAP_TOOLS_ON_INSTALL=1 npm install -g \
 
 
 class PrimeAgentHarnessConfig(HarnessConfig):
-    commit: Literal["514633727bf26d74f39f3119c2b0e31a5ceb2a9d"] = PRIME_AGENT_COMMIT
+    commit: Literal["81ae3cb34d27d38ee37f9e205a1e73694993b344"] = PRIME_AGENT_COMMIT
     """Prime Agent main commit to install."""
 
     autonomous: bool = False

verifiers/v1/harnesses/rlm/harness.py

@@ -47,10 +47,13 @@ class CompactionConfig(BaseConfig):
     summarize_at_tokens: PositiveInt | None = None
     """Compact at this token count. When unset, compact when 16k tokens remain below the
     model context window when the provider advertises it."""
+    max_compactions: PositiveInt | None = None
+    """Compactions per session before the engine stops compacting; `None` =
+    nano-rlm's default."""
 
 
 class RLMHarnessConfig(HarnessConfig):
-    version: str = Field(default="240090d", min_length=1)
+    version: str = Field(default="dd2c04f", min_length=1)
     """Git ref (branch, tag, or commit) of nano-rlm to install. Must know every
     field this harness puts on the wire, i.e. be at least the default ref."""
     max_depth: NonNegativeInt | None = None
@@ -59,8 +62,11 @@ class RLMHarnessConfig(HarnessConfig):
     builtin_skills: list[BuiltinSkill] = Field(default_factory=list)
     """Built-in rlm skills to enable (RLM_SKILLS), e.g. `["edit"]`; empty enables none.
     The tool set is fixed (ipython); the base `skills` field takes SKILL.md paths."""
-    compaction: CompactionConfig | None = None
-    """Context compaction policy. Set an empty config to use automatic thresholds."""
+    compaction: CompactionConfig | bool | None = None
+    """Context compaction: a `[compaction]` section (or `true`) enables it with the
+    given settings, `false` disables it (an overflowing session then fails), and
+    unset defers to nano-rlm's default (on, automatic thresholds, bounded by its
+    default 1M tree-token budget)."""
     max_concurrent_subagents: PositiveInt | None = None
     """Sub-agents running at once per session tree; `None` = nano-rlm's default (4).
     nano-rlm requires it to be at least `max_depth`."""
@@ -146,6 +152,19 @@ class RLMHarness(ACPHarness[RLMHarnessConfig]):
         system_prompt: str | None,
     ) -> JsonObject:
         compaction = self.config.compaction
+        policy_knobs: dict[str, Any] = {
+            "max_depth": self.config.max_depth,
+            "max_concurrent_subagents": self.config.max_concurrent_subagents,
+            "max_total_turns": self.config.max_total_turns,
+            "max_total_tokens": self.config.max_total_tokens,
+            "max_tool_output_bytes": self.config.max_tool_output_bytes,
+        }
+        if isinstance(compaction, bool):
+            policy_knobs["compaction"] = compaction
+        elif compaction is not None:
+            policy_knobs["compaction"] = True
+            policy_knobs["summarize_at_tokens"] = compaction.summarize_at_tokens
+            policy_knobs["max_compactions"] = compaction.max_compactions
         appends = [
             text
             for text in (system_prompt, self.config.append_to_system_prompt)
@@ -161,19 +180,7 @@ class RLMHarness(ACPHarness[RLMHarnessConfig]):
             # None = passthrough: the key stays off the wire and nano-rlm's own
             # default applies.
             "policy": {
-                key: value
-                for key, value in {
-                    "max_depth": self.config.max_depth,
-                    "compaction": compaction is not None,
-                    "summarize_at_tokens": (
-                        compaction.summarize_at_tokens if compaction else None
-                    ),
-                    "max_concurrent_subagents": self.config.max_concurrent_subagents,
-                    "max_total_turns": self.config.max_total_turns,
-                    "max_total_tokens": self.config.max_total_tokens,
-                    "max_tool_output_bytes": self.config.max_tool_output_bytes,
-                }.items()
-                if value is not None
+                key: value for key, value in policy_knobs.items() if value is not None
             },
             "system_prompt_path": None,
             "append_to_system_prompt": "\n\n".join(appends) or None,

{verifiers-0.3.2.dev37.dist-info → verifiers-0.3.2.dev39.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.5
 Name: verifiers
-Version: 0.3.2.dev37
+Version: 0.3.2.dev39
 Summary: Verifiers: Environments for LLM Reinforcement Learning
 Project-URL: Homepage, https://github.com/primeintellect-ai/verifiers
 Project-URL: Documentation, https://github.com/primeintellect-ai/verifiers

{verifiers-0.3.2.dev37.dist-info → verifiers-0.3.2.dev39.dist-info}/RECORD

@@ -107,9 +107,9 @@ verifiers/v1/harnesses/pi/harness.py,sha256=M0BRsfJxnBNP-KCrp0vT5gHlaJEDnyti5x_3
 verifiers/v1/harnesses/pool/__init__.py,sha256=HTYsNiWGdiNEsSHj4xhPmshJSNb7SHjQ7d4yZQ4p3HM,127
 verifiers/v1/harnesses/pool/harness.py,sha256=HKAogPbyLK5QpvPQ1teHOh_IyRQX-or2JF71nnZ_5Ps,3875
 verifiers/v1/harnesses/prime_agent/__init__.py,sha256=xfPC4RrlnyYWaf38tPAOZq4YL6M55lMpj5BoRLmaK8I,171
-verifiers/v1/harnesses/prime_agent/harness.py,sha256=byplbbIOFACCn2BvVZAxBSHdaPKBGERYz_hA1rgBj0E,11205
+verifiers/v1/harnesses/prime_agent/harness.py,sha256=_FJtiYvWOsBpjEBKuJsnmuldYNwJxKK8ONa5e-c_mYM,11205
 verifiers/v1/harnesses/rlm/__init__.py,sha256=hwx51xTdEWzzYgMj-p4ETNcptoub3IPMDbEdJ4uB8Ns,122
-verifiers/v1/harnesses/rlm/harness.py,sha256=uHP8VL2CCviZr65PF0IEfgpXxj0uph0E2t6YRvmF1zM,9845
+verifiers/v1/harnesses/rlm/harness.py,sha256=1a88tUM8l4BRf-In-dPyhtW9ilyH2GJdbh1ezL2BRcQ,10297
 verifiers/v1/harnesses/terminus_2/__init__.py,sha256=l1pwfg1HyVDEqT9efdnFWcXqaZwtKwe7TrplkyxZSA0,166
 verifiers/v1/harnesses/terminus_2/harness.py,sha256=kXKhi-aqqsoXGnvet0Tq_v7LYYpRYEzARyQL2MTEGWs,3101
 verifiers/v1/harnesses/terminus_2/program.py,sha256=X8MZdyItU3gbXGXnGOlJ9LWcUTnqIDaxlhkZvoRiqKY,2703
@@ -185,8 +185,8 @@ verifiers/v1/utils/prime.py,sha256=UTYRjp9cbjNb6CVmBHda-1wWZAIIfxNOmTyNuT7_wL4,9
 verifiers/v1/utils/retries.py,sha256=Y2ZgrAjn-qNkRKZP_RVNL_7EK0iaRcZeCa404lpGYi4,5417
 verifiers/v1/utils/score.py,sha256=493yJVMw8teCu9JxapxMFPFzI0hNUqdo0Y2nGW4kckk,6200
 verifiers/v1/utils/version.py,sha256=-obEo_-l9-D8FLef4hYxncOe-uJpxrM1g2Hig_37Sgs,1607
-verifiers-0.3.2.dev37.dist-info/METADATA,sha256=Kba9nRG0TlzRVxS9P51ScU0eYJJ8I4bRKDGEgMZuxY4,4065
-verifiers-0.3.2.dev37.dist-info/WHEEL,sha256=zOwg4jB6zX2kU910N-cMawjivD6tO8NEWvE12je1bVk,87
-verifiers-0.3.2.dev37.dist-info/entry_points.txt,sha256=uqQje0TMsr7k6nXlWxlPs13Si0evb3pVCEDF6c-YL80,241
-verifiers-0.3.2.dev37.dist-info/licenses/LICENSE,sha256=v0RrUsdV3IDoZhrRce297IXS3xMHNJ-_LdLpFAUWb9k,1072
-verifiers-0.3.2.dev37.dist-info/RECORD,,
+verifiers-0.3.2.dev39.dist-info/METADATA,sha256=aCN9__3iHq_uJT58Pv_szAvF_zQiUkj2j_IihmNAi3I,4065
+verifiers-0.3.2.dev39.dist-info/WHEEL,sha256=zOwg4jB6zX2kU910N-cMawjivD6tO8NEWvE12je1bVk,87
+verifiers-0.3.2.dev39.dist-info/entry_points.txt,sha256=uqQje0TMsr7k6nXlWxlPs13Si0evb3pVCEDF6c-YL80,241
+verifiers-0.3.2.dev39.dist-info/licenses/LICENSE,sha256=v0RrUsdV3IDoZhrRce297IXS3xMHNJ-_LdLpFAUWb9k,1072
+verifiers-0.3.2.dev39.dist-info/RECORD,,