Add Perplexity-enabled TerminalBench agent - #763

Draft

neubig wants to merge 17 commits into

mainfrom

agent/perplexity-cli-terminalbench

Commits on Jul 16, 2026

Support proxy credentials for Pier mini-SWE runs

neubig
and
openhands-agent
committed
d30e32c
Add bounded-history model for MiniSWE evaluations
neubig
committed
0770cbe

Commits on Jul 17, 2026

Honor verifier rewards after agent failures
neubig
committed
a50c041
Refresh Harbor results before evaluation
neubig
committed
5e8c811
Keep optional MiniSWE import type-checkable
neubig
committed
115bcdf

Commits on Jul 30, 2026

Add Perplexity-enabled TerminalBench agent

neubig
and
openhands-agent
committed
e53ac8a
Use resolvable Perplexity skill ref
neubig
committed
e9b7f79
Load Perplexity Harbor agent outside benchmarks package
neubig
committed
d940ac0

Commits on Jul 31, 2026

Require Perplexity search in TerminalBench runs
neubig
committed
37c38e7
Allow PPLX CLI API egress in TerminalBench
neubig
committed
e272429

Commits on Aug 3, 2026

Allow longer PPLX Harbor agent setup
neubig
committed
5b639f3

Commits on Aug 5, 2026

Pass TerminalBench Harbor runtime mounts
neubig
committed
a5712d1
Support mounted runtime in PPLX Harbor agent
neubig
committed
51916bf
Require PPLX bootstrap search in Harbor agent
neubig
committed
77e727d
Install PPLX CLI without task image dependencies
neubig
committed
80665e4
Use PPLX binary from mounted Harbor runtime
neubig
committed
29766e3
Checkpoint Harbor trials across Kubernetes retries
neubig
committed
ca38ea7